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

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

TRAUMA-RELATED TREATMENT GAINS AMONG WOMEN
WITH HISTORIES OF INTERPERSONAL VIOLENCE AND
CO-OCCURRING MENTAL HEALTH AND SUBSTANCE ABUSE DISORDERS
by

Margaret-Anne Mackintosh

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
(PSYCHOLOGY)

August 2009

Copyright 2009 Margaret-Anne Mackintosh
ii
Dedication

These papers are in memory of four wonderful family members and friends who
touched my life deeply and, unfortunately, passed away just before I realized this
dream; John M. Mackintosh Sr., Frances Z. Mackintosh, Norma G. Hammond, and
Kevin Corker, Ph.D. My father blessed me with intellectual curiosity and a passion for
learning, my mother modeled determination and provided unwavering support
throughout my life, Norma was a spiritual mother and stoked my passion for service to
others, and Kevin was my model of a true scientist and a good soul who embraced life.

iii
Acknowledgments
There are a great many to thank for their help in completing this project. First of
all, to my advisor, Margy Gatz, Ph.D., without your mentorship, support, patience, and
all the amazing opportunities you provide your students, I would not be in this position.
I would like to thank the other faculty who served on my dissertation committee (Drs.
Gayla Margolin, J. Jack McArdle, Karen Hennigan, and Penelope Trickett) for their
guidance, insight, and support for my work. I am indebted to the “Study of Women”
staff for their amazing job conducting the complex study on which my dissertation was
based. I would especially like to thank the study’s principal investigators, Drs. Margy
Gatz, Karen Hennigan, and Maura O’Keefe, and the project coordinator, Dr. Tara Rose,
for their hard work on the project and for my opportunity to be a part of it. I am deeply
indebted to Dr. Archana Jajodia for guidance on conducting some of the statistical
analyses and her aid in getting me going again. I thank the other students, past and
present, in the Section on Clinical Research in Aging and Psychology (SCRAP lab) and
my amazing cohort at USC for their help, encouragement, thoughtful editing, and input
throughout my doctoral work, especially Poorni G. Otilingam, M.P.H., M.A., Dr. Lara
Heflin, and Dr. Michelle Yeh for their help and walking with me through illness and
discouragement. Also, my deep thanks go to James, Rebecca, William and Grace
Hiramoto for their long-time support and quiet space to finish up my papers. Last, and
definitely not least, for the women who participated in this study; I am deeply grateful
for their willingness to share their lives with us, and appreciate how much they taught
me about facing challenges.
iv
Table of Contents
Dedication..................................................................................................................... ii
Acknowledgments........................................................................................................ iii
List of Tables .................................................................................................................v
List of Figures............................................................................................................. vii
Abstract...................................................................................................................... viii
Chapter 1: General Introduction ...................................................................................1
Chapter 2: Longitudinal consistency and factor structure of the PSS-SR
among women with co-occurring mental health disorders and histories of
interpersonal trauma.......................................................................................................6
Chapter 2: Abstract .................................................................................................6
Chapter 2: Introduction...........................................................................................7
Chapter 2: Results.................................................................................................19
Chapter 2: Discussion ...........................................................................................22
Chapter 3: Impact of trauma-focused treatment on trauma-related outcomes
among women with co-occurring disorders and histories of interpersonal
violence........................................................................................................................29
Chapter 3: Abstract ...............................................................................................29
Chapter 3: Introduction.........................................................................................30
Chapter 3: Method ................................................................................................47
Chapter 3: Results.................................................................................................61
Chapter 3: Discussion .........................................................................................116
References..................................................................................................................133
Appendix A ...............................................................................................................150
Appendix B ................................................................................................................155
Appendix C ................................................................................................................157

v
List of Tables
Table 2.1. PSS-SR Items (Mapped To DSM-IV Criteria) and Factor Structures
for Each Model Tested.................................................................................................10
Table 2.2. Modeling Results for Unconstrained Longitudinal Factorial Analyses
of Posttraumatic Stress Symptoms...............................................................................20
Table 3.1. Means (and SD) for Study Variables by Treatment Group. ......................62
Table 3.2. Means (SD and sample sizes) for Study Variables by Interview
Period and Treatment Group........................................................................................65
Table 3.3. Longitudinal Latent Difference Score Models of Posttraumatic
Stress Symptoms..........................................................................................................70
Table 3.4. Means of Difference Scores for Posttraumatic Stress Symptoms by
Treatment Group, Interview Period, and Seeking Safety Classes. ..............................74
Table 3.5. Means of Difference Scores for Posttraumatic Stress Symptoms by
Treatment Group, Interview Period, and Use of Residence Services Group...............77
Table 3.6. Longitudinal Latent Difference Score Models of Unsafe Events..............80
Table 3.7. Means of Difference Scores for Unsafe Events by Treatment Group,
Interview Period and Seeking Safety Classes..............................................................85
Table 3.8. Means of Difference Scores for Unsafe Events by Treatment Group,
Interview Period, and Use of Residential Services Group...........................................86
Table 3.9. Model Summary Table for Longitudinal Latent Difference Score
Models of Days in Residential Treatment ...................................................................88
Table 3.10. Raw Means for Days in Residential Treatment Settings by
Treatment Group, Interview Period, and Seeking Safety Attendance.........................96
Table 3.11. Raw Means for Days in Residential Treatment Settings by
Treatment Group, Interview Period, and Seeking Safety Attendance.........................97
Table 3.12a: Modeling Results for Final Model: Latent Growth Curve Model
Parameter Estimates...................................................................................................103
Table 3.12b: Modeling Results for Final Model: Cross-lagged Associations
Parameter Estimates...................................................................................................104
vi
Table 3.12c: Modeling Results for Final Model: Parameter Estimates for
Effects of Treatment Group on Dependent Variables................................................105
Table 3.12d: Modeling Results for Final Model: Parameter Estimates for
Effects of Number of Seeking Safety Classes Taken during Months One through
Three on Dependent Variables...................................................................................106
Table 3.12e: Modeling Results for Final Model: Parameter Estimates for
Effects of Number of Seeking Safety Classes Taken during Months Four to Six
on Dependent Variables.............................................................................................107
Table 3.12f. Modeling Results for Final Model: Correlated Growth Factors. .........108
Table A-1. Description of Seeking Safety Sessions Implemented at Los
Angeles Site of WCDVS ...........................................................................................150
Table B-1. PSS-SR Items by Subscale .....................................................................155
Table C-1. Items on Unsafe Events Scale.................................................................157

vii
List of Figures
Figure 2.1. Scatter plot with ordinary least squares regression line depicting fit
of modeling results for structural models of posttraumatic stress symptoms..............21
Figure 3.1. Outline of study enrollment, retention, and trauma-focused
treatment ......................................................................................................................48
Figure 3.2. Schematic of dual change model for posttraumatic stress symptoms. .....59
Figure 3.3. Posttraumatic stress symptom scores across time for 40 random
cases .............................................................................................................................64
Figure 3.4. Number of different types of unsafe events for 40 random cases ............64
Figure 3.5 Days in residential treatment for a random sample of 40 cases ................67
Figure 3.6. Final univariate longitudinal model of posttraumatic stress
symptoms .....................................................................................................................69
Figure 3.7. Depiction of partial modeling results of final model for
posttraumatic stress symptoms, Treatment Group, and Seeking Safety
attendance ....................................................................................................................73
Figure 3.8. Final modeling results for unsafe events..................................................78
Figure 3.9. Partial modeling results for unsafe events, treatment group, and
Seeking Safety attendance ...........................................................................................84
Figure 3.10. Final modeling results for days in residential treatment ........................91
Figure 3.11. Partial modeling results for days in residential treatment with
Treatment Group and Seeking Safety attendance........................................................93

viii
Abstract
Interpersonal violence directed against girls and women is both widespread and
can lead to serious long-term consequences, including the development of co-occurring
mental health and substance abuse disorders. This two-part study investigated
psychosocial gains following trauma-focused treatment among women with histories of
interpersonal violence and co-occurring disorders. The data are drawn from the Los
Angeles site of the national Women, Comorbid Disorders and Violence Study
(WCDVS), which assessed the effects of integrated substance abuse and mental health
treatment using an intent-to-treat quasi-experimental design. A diverse sample of 370
women was interviewed up to five times over 12-months. The first part of the present
study established the longitudinal consistency of a measure of posttraumatic stress
symptoms. Results indicated that five of the 17 items performed inconsistently across
time and were dropped. The second part investigated the impact of a trauma-focused
treatment program (Seeking Safety; Najavits, 2002) on posttraumatic stress symptoms
and unsafe events. The impact of attendance on outcome measures was assessed using
longitudinal structural equation models, and statistically adjusted for days in residential
treatment and WCDVS treatment group. Results indicated that while there were
significant decreases in posttraumatic stress symptoms across time, level of trauma-
focused treatment did not predict these changes. Significant reductions in posttraumatic
stress symptoms were predicted by greater use of residential treatment services early in
the woman’s enrollment in the study. Women in both treatment conditions showed
significant reductions in unsafe events at six months. For women in the integrated
ix
services group, greater exposure to trauma-focused treatment was associated with fewer
unsafe events. Finally, greater participation in trauma-focused treatment predicted
greater use of residential services, even after women had completed the trauma-focused
treatment program. While the reduction in unsafe events suggests increased use of
safety behaviors and coping skills like those taught in Seeking Safety, the reductions in
posttraumatic stress symptoms across different residential treatment programs was
unexpected. Increased awareness of the negative impact of traumatic experiences and
availability of trauma-informed service providers may lead to dissemination of trauma-
relevant information into the general treatment programs.
1
Chapter 1: General Introduction
The impact of interpersonal violence, especially that occurring during childhood
and adolescence, can have long-term and widespread impact on the lives of those
affected. Women are more vulnerable to interpersonal types of psychologically
traumatic experiences (e.g., sexual assault, physical violence within relationships, and
childhood violence) compared to men (Kessler et al., 1999). Not only are they more
often the victims of interpersonal violence, but victimization entails increased risk for
serious mental health outcomes, such as recurrent depression, substance abuse
disorders, emotional dysregulation, suicidality and self-harm behaviors (van der Kolk,
McFarlane, & Weisæth, 1996). In addition to negative psychological outcomes, long-
term consequences of poor adaptation following traumatic exposure include physical
health problems, difficulties with parenting, poor educational and vocational attainment,
and increased risk for incarceration and repeated incarcerations, leading to an ever-
increasing set of difficulties for individuals struggling with the after effects of
psychological trauma (Becker et al., 2005; Kirmayer, Lemelson, & Barad, 2007;
Schnurr & Green, 2004; Terr, 1991).
At the same time, intervention has been inadequate. The complex and
sometimes fundamental deficits in emotional processing and interpersonal skills found
among women with histories of interpersonal trauma, mental health and substance
abuse disorders influence poor treatment outcomes (Linehan, Bohus, & Lynch, 2007;
Mueser, Noordsy, Drake, & Fox, 2003; Najavits, 2002). In addition, treatment facilities
often lack the resources and training to provide the range of services women need to
2
overcome mental health, substance abuse, educational, vocational, legal, and medical
challenges they often face as trauma survivors (Elliott, Bjelajac, Fallot, Markoff, &
Reed, 2005; Harris, 1994; Holdcraft & Comtois, 2002; Salasin, 2005). Research has
shown that women with comorbid disorders, in fact, are not likely to receive care for
both disorders, have poorer treatment outcomes, present with more complex cases, and
are frequent consumers of services without significant long-term changes in functioning
(P. J. Brown, Recupero, & Stout, 1995; Drake, O'Neal, & Wallach, 2008; Najavits,
2003; Timko & Moos, 2002).
Fortunately, a number of factors appear to be changing, for the better, for this
group, such as 1) consumer-survivor-recovery movement has highlighted the need for
better and integrated services and has involved service users in the design and
implementation of their treatment programs; 2) increased use of trauma-informed
services, integrated mental health and substance abuse treatment programs, and
increased use of wrap-around services; 3) funding and development of research
programs testing previously held assumptions about this population (e.g., you cannot
treat both substance abuse and PTSD at the same time); and 4) development of
manualized treatment programs that provide flexibility in implementation while
addressing specific skill deficits and needs to specific subgroups (Hendrickson, Schmal,
& Ekleberry, 2004; Mueser et al., 2003; Salasin, 2005).
The two papers presented here take advantage of many of these advancements in
the treatment of women with co-occurring mental health and substance abuse disorders
who also have a history of interpersonal violence. Both use data from the Los Angeles
3
site (Gatz, Hennigan, O'Keefe, & Rose, 2004) of the Women, Comorbid Disorders and
Violence Study (McHugo, Kammerer et al., 2005), which was one of the first studies
investigating the impact of trauma-informed treatment programs that integrated
treatment for both mental health and substance abuse services. The WCDVS found
small but significant improvements in mental health and posttraumatic stress symptoms,
especially when more integrated services were offered to women (Ellis & Morrissey,
2009; Gatz et al., 2007; Morrissey, Jackson et al., 2005). The two papers presented here
provide a more detailed test of specific aspects of these broader findings.
The first paper details an investigation of the psychometric properties of the
measure of posttraumatic stress symptoms used in the WCDVS, while the second
documents the impact of a trauma-focused treatment protocol designed to teach safety-
enhancing skills among individuals with PTSD and substance abuse disorders. This
pair of papers demonstrates the importance of clear development and understanding of
psychological constructs before taking them into the more practical world of a treatment
outcome study in a realistic environment. These papers detail a balance or dance
between concerns about consistency and validity of measurement (enhancing internal
validity) while taking advantage of an outcome study steeped in the intricacies of
providing psychological treatment in real-world settings (enhancing external validity).
The first paper establishes a longitudinally consistent measure of posttraumatic stress
symptoms, which has the benefit of being able to attribute changes across the study
period to true changes in measured symptoms rather than to either measurement
inconsistency, specific factors variance or a combination of these. In line with previous
4
studies of trauma-focused treatment, this global measure of posttraumatic stress
symptoms is used in the second part of the study. However, the first paper also explores
the underlying factor structure of posttraumatic stress symptoms. These types of
analyses provide insight into the relationship among the various factors proposed to
underlie the construct of PTSD, specifically arousal, avoidance, numbing, and re-
experiencing symptoms, and how these symptom clusters are related to one another and
what might maintain posttraumatic stress symptoms across time.
The second paper documents the effects of the integrated PTSD and substance
abuse disorders treatment program from the Los Angeles WCDVS site. The trauma-
focused treatment protocol, Seeking Safety (Najavits, 2002) focuses on teaching women
better safety skills, including cognitive, behavioral and interpersonal coping and
emotion regulation skills. This study used a quasi-experimental, intent-to-treat design
that followed women across a 12-month period, assessing various psychological,
physical, social and service use variables every three or six months. While all the
women in this study started out in either the intervention group (integrated mental
health and substance abuse services, subsequently referred to as the Integrated
Treatment Group) or the comparison group (receiving primarily substance abuse
services, subsequently referred to as the Treatment-As-Usual Group), they were not
randomly assigned to these conditions, leaving concerns about possible differences
between groups based on treatment program. However, the groups do accurately
represent the types of women showing up at individual treatment settings. Previous
publications (Gatz et al., 2007; Gatz et al., 2004) investigated many factors and found
5
few differences between the groups at baseline (see Paper 2, Section 2.3, for a further
discussion). Moreover, the women were followed across time, wherever they went
regardless of whether they remained in treatment or not. While these real world factors
make clean hypothesis testing more difficult, Paper 2 uses longitudinal structural
equation modeling to investigate the interrelationships among the study variables --
posttraumatic stress symptoms, unsafe events, and days in residential treatment -- with
the focus on assessing the impact of participation in Seeking Safety on outcome
measures.
6
Chapter 2: Longitudinal consistency and factor structure of the PSS-SR
among women with co-occurring mental health disorders and histories of
interpersonal trauma

Chapter 2: Abstract
With the increasing use of longitudinal research designs, establishing that a
measure assesses constructs consistently across time, as well as performing well at each
individual time period, has become increasingly important. We examined the
longitudinal consistency and factor structure of the PTSD Symptom Scale – Self Report
in a diverse sample of women with co-occurring substance abuse and mental health
disorders and histories of interpersonal trauma. Data were from the Los Angeles site of
the Women, Comorbid Disorders and Violence Study. Symptom reports from 370
women enrolled in residential treatment were collected at baseline and every three
months for one year. Results indicated that five of the 17 items behaved inconsistently
across time and were dropped. Eight factor structures were tested. The best-fitting
factor structure that could be modeled across time included two factors: 1) Re-
experiencing/Arousal and 2) Numbing/Avoidance. The modified scale allows for
clearer interpretations of changes across time, and establishes its psychometric
properties in a new population.
7
Chapter 2: Introduction
Researchers and clinicians often wish to study the consequences of trauma
longitudinally, whether it is the development of psychological effects of trauma after
initial exposure or changes in posttraumatic stress symptoms during a treatment
episode. When conducting longitudinal research, clinicians often implicitly assume that
their instruments are consistently measuring the same construct each time the person is
assessed, or longitudinal factorial invariance (Horn & McArdle, 1992). Without such
consistency, any differences in scores from pre-treatment to post-treatment cannot be
clearly interpreted. In other words, differences on a measure across time could reflect
the true impact of time on posttraumatic stress symptoms, inconsistencies in the
measurement of posttraumatic stress symptoms, or a combination of both; resulting in
non-interpretable or misinterpreted research results. With the advances in psychometric
procedures and increased accessibility of these techniques through modern statistics
programs, the need to assess important psychometric properties of our measures cannot
be overlooked.
One frequently used measure of posttraumatic stress symptoms is the PTSD
Symptom Scale – Self Report (PSS-SR, Foa et al., 1993). Only one study to date has
examined the longitudinal factorial invariance of this measure (Baschnagel, O'Connor,
Colder, & Hawk, 2005). These authors found that while the factor structure was the
same between two measurement periods two months apart, the factor loadings were not
equal across time. This result suggests that the same symptom clusters were present
across time; however, the relationship between each symptom and its associated
8
symptom cluster was not consistent. Thus, the meaning of differences in symptom
scores across time is difficult to interpret.
There have been especially few studies of the validity and reliability of
posttraumatic stress symptoms measures among substance abusing individuals (Coffey,
Dansky, Falsetti, Saladin, & Brady, 1998; Stewart, Conrod, Pihl, & Dongier, 1999). No
studies have investigated posttraumatic stress symptom scale functioning among those
with substance abuse disorders as well as another mental health disorder (i.e., co-
occurring disorders). Because women with PTSD are 2.5 to 4.5 times more likely to
have a substance use disorder compared to women without PTSD (Chilcoat & Menard,
2003), this is an important subpopulation of individuals to study. Thus, this study
investigated whether the PSS-SR consistently measures posttraumatic stress symptoms
across time among women with co-occurring disorders and a history of interpersonal
trauma.
Another important issue when assessing the utility of a scale is establishing the
organizational structure of the construct. A consistent factor structure for posttraumatic
stress symptoms and Posttraumatic Stress Disorder (PTSD) is still being investigated
(McWilliams, Cox, & Asmundson, 2005; Palmieri & Fitzgerald, 2005; Simms, Watson,
& Doebbeling, 2002). Factor structures based on diagnostic criteria (American
Psychiatric Association, 1994) have not been found to be the best fitting models in
psychometric studies (Buckley, Blanchard, & Hickling, 1998; Witteveen et al., 2006).
Drawing on both current theoretical models of PTSD and previous research studies on
the factor structure of the PSS-SR at single time points, this study tested eight factorial
9
models of posttraumatic stress symptoms longitudinally. Table 2.1 describes the 17
items from the PSS-SR and the eight hypothetical models tested.
In Model 1 all 17 symptoms load on one global factor. This is the most
parsimonious model possible and thus provides a test of the unitary construct PTSD.
This model has found some support in the literature (D. W. King, King, Fairbank,
Schlenger, & Surface, 1993), although probably is best thought of as a comparison
model.
Model 2 consists of two intercorrelated latent variables representing
theoretically reciprocating processes of 1) Re-experiencing/Effortful Avoidance and 2)
Numbing/Arousal (Feuer, Nishith, & Resick, 2005; Horowitz, 1986). This model has
been supported in three studies (Buckley et al., 1998; Stewart et al., 1999; Taylor, Kuch,
Koch, Crockett, & Passey, 1998). However, in Buckley et al. (1998) the two arousal
items classically associated with PTSD (i.e., hypervigilance and exaggerated startle
response) loaded on the Re-experiencing/Effortful Avoidance factor, leaving the
remaining symptoms on the second factor appearing much like depression.
Model 3 consists of the three factors suggested in the DSM-IV, i.e., Re-
experiencing, Effortful Avoidance/Numbing, and Arousal (American Psychiatric
Association, 1994) subsumed under the broader factor of PTSD. Several studies have
failed to support this model psychometrically (DuHamel et al., 2004; Foa, Riggs, &
Gershuny, 1995; D. W. King, Leskin, King, & Weathers, 1998; L. A. King & King,
1994; Palmieri & Fitzgerald, 2005; Stewart et al., 1999). One study found this model to

10
Table 2.1. PSS-SR Items (Mapped To DSM-IV Criteria) and Factor Structures for Each Model Tested

Item topic

Model 1 Model 2a Model 3a Models 4a/b Model 5 Models 6a/b
1 (B1). Intrusive thoughts

PTSD R/Av R R R R
2 (B2). Nightmares

PTSD R/Av R R R R
3 (B3). Reliving trauma

PTSD R/Av R R R R
4 (B4). Emotional cue reactivity

PTSD R/Av R R R R
5 (B5). Physiological cue reactivity

PTSD R/Av R R R R
6 (C1). Avoidance of thoughts/feelings

PTSD R/Av Av/N Av Av Av
7 (C2). Avoidance of reminders

PTSD R/Av Av/N Av Av Av
8 (C3). Limited recall of event

PTSD N/Ar Av/N N Gd N
9 (C4). Loss of interest in significant
activities

PTSD N/Ar Av/N N Gd N

PTSD = Single PTSD Factor; R = Re-experiencing Factor; Av = Avoidance Factor; N = Numbing Factor; Ar = Arousal
Factor; Gd = General Distress Factor; Ga = General Arousal Factor; H = Hyperarousal Factor. Models with suffix a
assume intercorrelated symptom clusters; Models with suffix b are hierarchical models that assume a higher order
factor subsuming symptom clusters.

11
Table 2.1. Continued.

Item topic

Model 1 Model 2a Model 3a Models 4a/b Model 5 Models 6a/b
10 (C5). Detachment from others

PTSD N/Ar Av/N N Gd N
11 (C6). Emotional numbness

PTSD N/Ar Av/N N Gd N
12 (C7). Foreshortened sense of future

PTSD N/Ar Av/N N Gd N
13 (D3). Concentration problems

PTSD N/Ar Ar Ar Gd Ga
14 (D2). Irritability or anger

PTSD N/Ar Ar Ar Gd Ga
15 (D1). Sleep problems

PTSD N/Ar Ar Ar Gd Ga
16 (D4). Hypervigilance

PTSD N/Ar Ar Ar Ar H
17 (D5). Exaggerated startle response

PTSD N/Ar Ar Ar Ar H

Notes: PTSD = Single PTSD Factor; R = Re-experiencing Factor; Av = Avoidance Factor; N = Numbing Factor; Ar =
Arousal Factor; Gd = General Distress Factor; Ga = General Arousal Factor; H = Hyperarousal Factor. Models with
suffix a assume intercorrelated symptom clusters; Models with suffix b are hierarchical models that assume a higher
order factor subsuming symptom clusters.
12

fit better than a one-factor solution, although no additional models were tested
(Cordova, Studts, Hann, Jacobsen, & Andrykowski, 2000).
Models 4a and 4b posit four symptom clusters: Re-experiencing, Effortful
Avoidance, Numbing and Arousal. Several studies have supported this disaggregation
of the DSM-IV’s Effortful Avoidance/Numbing cluster (Asmundson, Wright,
McCreary, & Pedlar, 2003; Foa et al., 1995; D. W. King et al., 1998; L. A. King &
King, 1994; Marshall, 2004; Palmieri & Fitzgerald, 2005; Shelby, Golden-Kreutz, &
Andersen, 2005). In Model 4a, the four correlated factors are not subsumed under a
single higher-order factor (e.g., PTSD), suggesting that the individual symptom clusters
may have different underlying causes rather than a single cause (Marshall, 2004). In
Model 4b the four subscales are included under a single higher-order PTSD factor. This
model has been supported in three studies (Andrews, Joseph, Shevlin, & Troop, 2006;
Asmundson et al., 2000; L. A. King & King, 1994). The hierarchical model is an
important theoretical and conceptual model as it suggests that the four symptom
clusters, while individual processes, are still part of a larger construct of PTSD.
Model 5 is a four-factor model based on the work of Simms et al. (2002) in
which the prototypic hyperarousal symptoms (i.e. hypervigilance and exaggerated
startle response) were disaggregated from the other Arousal symptoms (i.e. sleep
problems, irritability/outbursts of anger, and concentration problems). These three
remaining symptoms were considered part of a dysphoria or general distress factor that
also included the Numbing factor. The division of Arousal symptoms into two groups
13

is supported by one recent study investigating the factor structure of the German version
of the PSS-SR (Griesel, Wessa, & Flor, 2006). Griesel and her colleagues (2006) found
a three factor solution consisting of 1) Re-experiencing/Effortful Avoidance; 2)
Numbing/Arousal; and 3) a second Arousal cluster containing only the hyperarousal
items. The four-factor model based on Simms et al., (2002) also was found to be the
best fitting model factor structure by Baschnagel et al. (2005) while other results have
supported it (Palmieri & Fitzgerald, 2005).
Two final models were added by the authors. Models 6a and 6b are five factor
models including the Re-experiencing, Effortful Avoidance, Numbing, and the two
Arousal symptom clusters from Models 4a and 4b. The Arousal cluster was divided
into two separate clusters representing the prototypic hyperarousal symptoms first
separated by Simms and his colleagues (2002) and the general arousal symptoms
common to many anxiety disorders and depression. As the Numbing symptom cluster
has theoretical and research importance (Flack, Litz, Hsieh, & Kaloupek, 2000; Litz et
al., 1997), it was retained as a separate factor. Models 6a and 6b are the same except
Model 6b includes the higher order PTSD factor.
As only one other study on the longitudinal assessment of the factor structure for
the PSS-SR was identified in the literature (Baschnagel et al., 2005), this study moves
the field forward not only by providing a replication of this previous work, but also by
extending the research by assessing item responses at more than two time points ,
testing the scale among an important population not previously studied: traumatized
14

women with co-occurring mental health disorders, and testing a greater number of
relevant models.
15

Chapter 2: Method
Data are from the Los Angeles site of the Women, Comorbid Disorder and
Violence Study (WCDVS). WCDVS was a national, multi-site study aimed at
developing and testing comprehensive treatment programs for women with co-
occurring mental health disorders, exposure to interpersonal trauma, and multiple
previous treatment episodes of either mental health or substance abuse services. For a
complete description of the study, please see McHugo et al. (2005). Each site included
an intervention group who received integrated mental health and substance programs
and a comparison group who received treatment as usual. Women were assessed at five
time points: baseline, 3-months, 6-months, 9-months, and 12-months. Multiple
measures were included at each time point, including the PSS-SR.
Participants
The Los Angeles site of the WCDVS included three residential treatment
agencies, one providing integrated care and two providing treatment-as-usual.
Participants were 370 women who completed a baseline interview, which occurred
within 30 days of entering one of the treatment agencies. Women’s average age was
33.2 years (SD = 8.7). The sample was racially diverse: 25.7% Hispanic, 35.9%
Caucasian, 23.2% Black, 0.8% Asian or Pacific Islander, 12.2% American Indian and
2.2% multiracial or listed race as “other”. Women had completed an average of 11.6
years of education (SD = 2.2). At baseline, each woman reported experiencing an
average of 16.0 different types of traumatic or highly stressful events (SD = 4.5) during
16

their lifetimes. Almost three quarters (73.9%) reported being victims of some form of
child abuse. Valid responses on the posttraumatic stress symptoms measure were
available from 356 women at baseline, 268 at 3-months, 276 at 6-months, 240 at 9-
months, and 300 at 12-months, with each woman providing an average of 3.8 valid
responses.
Measure
PTSD Symptom Scale – Self Report (PSS-SR). The PSS-SR is a 17 item
measure consisting of the 17 cardinal symptoms of PTSD (Foa et al., 1993). Each
symptom is rated on a 4-point scale, based on how bothered a person was by that
symptom during the last 30 days. Anchors were not at all/only one time (0) to 5 or
more times per week/almost always (3). Scale reliability for this sample, as measured
by Cronbach’s alpha, was .90 at baseline, .92 at 3-months, .89 at 6-months, .92 at 9-
months, and .93 at 12-months. Mean total scale scores were 15.3 at baseline, 13.8 at 3-
months, 19.9 at 6-months, 15.0 at 9-months, and 14.5 at 12-months, indicating moderate
levels of posttraumatic stress at each time period. See Appendix A for list of items.
Data Analyses
Assessment of model fit for structural models. The first step of data analysis
entailed establishing a factor structure for the PSS-SR. The eight models presented in
Table 2.1 were assessed, using data from all five time points. Each item could load on
one and only one factor, and this factor structure was the same across all five time
periods; however, the factor loadings were free to vary across time points (i.e.,
17

configural invariance, Horn & McArdle, 1992). In all the models described in this
paper, the total variability of scores across time was partitioned into between-person
variance and within-person variance. These represented the stable average scores for
each person and the unstable variation, due to individual specific variability and time-
varying measurement error, respectively (McArdle, Fisher, & Kadlec, in press).
Basic model fit was assessed via a number of standard model indices: each
model’s chi square value ( χ
2
) and its associated degrees of freedom, Akaike Information
Criterion (AIC, Akaike, 1973), Bayesian Information Criterion (BIC, Schwarz, 1978),
Comparative Fit Index (CFI, Bentler, 1990a), and root mean square error of
approximation (RMSEA, Steiger, 1990). With AIC and BIC there is no predetermined
value to identify better versus worse fitting models, however smaller values are better.
A CFI of at least .90 indicates adequate model fit (Bentler, 1990a). RMSEA values of
.05 or smaller indicate good-fitting models (Hu & Bentler, 1999).
While fit indices provided information about overall model functioning, we used
the penalty line method (McArdle & Nesselroade, 1994) to select the final structural
model. A penalty line was developed by regressing each model’s degrees of freedom
on the model’s χ
2
value, and the resulting regression line was plotted. The models that
are located at a larger vertical distance above the penalty line are poorer fitting models
and those with the largest vertical distance below the penalty line are the better fitting
models. Models that were the farthest below the penalty line were selected as the best
fitting, i.e. had the largest standardized regression residuals. This method is useful
18

because it balances model parsimony (i.e., fewer estimated parameters/more degrees of
freedom) with the model’s ability to reproduce the data (i.e., lower chi square values).
Assessment of longitudinal factorial invariance. Once the longitudinal factor
structure under conditions of configural invariance was determined, longitudinal
factorial invariance was further assessed by comparing models estimated under
conditions of configural invariance with models testing for metric invariance. In
metric invariant models, each item also was mathematically restricted to have equal
factor loadings across all five time periods. Because the more constrained metric
invariance models were nested within the models of configural invariance, differences
in χ
2
values ( Δχ
2
) were used to assess statistically significant differences between
models. Significant differences between models would indicate that women were
responding to items in different ways at different times. In this paper, items with factor
loadings that varied significantly statistically across time were dropped. The
assessment of the factor structure and the factorial invariance of the PSS-SR was
conducted using MPlus 4.0 (Muthén & Muthén, 2007) and used all available data.
Regression analyses were conducted using SPSS 11.5 (SPSS, 2002).
19

Chapter 2: Results
Assessment of model fit for structural models
Results from estimating the eight models described earlier are presented in
Table 2.2. Inspection of model fit indices suggests that all of the models fit the data
well except for Model 1, the single factor model.
When using the penalty line method that takes into account the number of
degrees of freedom in a model in relation to the model fit, different models are
highlighted as better fitting. Figure 2.1 depicts the scatter plot of each model’s χ
2
value
by its associated degrees of freedom as well as the penalty line. Points representing
Models 1 and 6a are above the penalty line, indicating poor model fit, while points for
all other models fall below the penalty line, indicating better model fit. Standardized
residuals for the four best fitting models are Model 4b = -.73, Model 2 = -.73, Model 6b
= -.63, and Model 3 = -.59. Since there was no difference in standardized residuals
between Model 4b and Model 2, both models were selected as the best-fitting model,
although Models 6b and 3 remain “respectable” alternatives. These analyses were
repeated using missing data estimation procedures and the results were very similar
with the same models being selected for further analysis.
Assessment of longitudinal factorial invariance
It proved impossible to test for the longitudinal metric invariance of Model 4b
because item 7 was dropped due to lack of invariance over time, leaving only one item
20

on the Effortful Avoidance factor. Thus, we could only assess Model 2 for metric
invariance across time. Tests indicated that factor loadings for five items varied

Table 2.2. Modeling Results for Unconstrained Longitudinal Factorial Analyses of
Posttraumatic Stress Symptoms

Model χ
2
df AIC BIC CFI RMSEA
1 1711 239 60078 60518 .83 .07
2 1051 237 59422 59873 .91 .05
3 885 23359263 59735 .92 .05
4a 703 227 59093 59596 .95 .04
4b 756 231 59139 59621 .94 .04
5 717 22759107 59611 .94 .04
6a 585 220 58989 59529 .96 .03
6b 679 229 59066 59558 .95 .04

Notes. df = degrees of freedom; AIC = Akaike’s Information Criterion; BIC =
Bayesian Information Criterion; CFI = Comparative Fit Index; RMSEA = Root Mean
Square Error of Approximation.

significantly across time and, the items were dropped: item 2, Δχ
2
(1) = 17.1; item 3,
Δχ
2
(1) = 7.4; item 5, Δχ
2
(1) = 7.1; item 7, Δχ
2
(1) = 7.8; and item 17, Δχ
2
(1) = 3.9; all
ps < .05. Items dropped were four Re-experiencing/Effortful Avoidance items (i.e.,
nightmares; reliving trauma; physiological cue reactivity; and avoidance of reminders),
21

Figure 2.1. Scatter plot with ordinary least squares regression line depicting fit of
modeling results for structural models of posttraumatic stress symptoms

and one Numbing/Arousal item (i.e., exaggerated startle response). After dropping
these items, the final two-factor model fit the data well; χ
2
(117) = 417, AIC = 43450,
BIC = 43718, CFI = .95, and RMSEA = .04. We also assessed the metric invariance for
the other models with respectable fit from the first step. Model 6b could not be tested
because, like Model 4b, item 7 needed to be dropped, leaving a one item factor. Eight
items in Model 3 did not perform consistently across time and were dropped. These
were items 2, 3, and 5 on the Re-experiencing factor and items 7, 9, 10, 11, and 12 on
the Avoidance/Numbing factor. This left only two items on each of these factors;
however all items were retained on the Arousal factor in this model.
22

Chapter 2: Discussion
This study investigated alternative models of PSS-SR’s factor structure and the
longitudinal consistency of the measure among women with co-occurring mental health
disorders and histories of interpersonal violence. We assessed the underlying structural
model of posttraumatic stress symptoms by testing six models of posttraumatic
symptoms often reported in the literature as well as two additional models proposed the
authors. The “best-fitting” longitudinal model depended on the criteria used and the
presence of sufficient items per factor to weather dropping items that did not perform
consistently across time.
Using a model selection technique that balanced model fit with the number of
parameters estimated (i.e., penalty line method, McArdle & Nesselroade, 1994), we
found that two other models fit the data equally as well. The first was a four-factor,
hierarchical model (Model 4b) consisting of separate factors of Re-experiencing,
Effortful Avoidance, Numbing, and Arousal items, all subsumed under a global PTSD
factor . The hierarchical structure suggests that the subscales are all related to a larger
unifying construct. This is in contrast to intercorrelated models, which conceptually
suggest that the underlying constructs are not necessarily caused by the same underlying
process. Two other studies have supported a similar model (Andrews et al,, 2006;
Asmundson et al., 2000; L. A. King & King, 1994) although many studies have found
that an intercorrelated four-factor model fits the data better than the hierarchical version
(Asmundson et al., 2003; DuHamel et al., 2004; Foa et al., 1995; D. W. King et al.,
23

1998; L. A. King & King, 1994; Marshall, 2004; Palmieri & Fitzgerald, 2005; Shelby et
al., 2005; Stewart et al., 1999). The second best-fitting model was a two-factor
intercorrelated model (Model 2) comprised of Re-experiencing/Effortful Avoidance and
Numbing/Arousal factors. Two other studies have found support for this structure
(Stewart et al., 1999; Taylor et al., 1998).
The second step of data analysis tested the longitudinal consistency or metric
invariance of the PSS-SR, i.e., whether the strength of the relationship between each
item and its common factor was the same at each time point. We attempted to test both
best-fitting models from the first step. However, only the two factor model had
sufficient time-invariant items per factor to test. Analyses testing each item separately
indicated that four items on the Re-experiencing/Effortful Avoidance factor and one
item from the Numbing/Arousal factor functioned differently across the five time
points. In other words, how women reacted to these items, in the context of the whole
scale, differed across time.
In the only other longitudinal study of the PSS-SR identified, Baschnagel and
his colleagues (2005) found that although the factor structure for the PSS-SR was
similar across a 2-month period, the factor loadings as a group were not equal across
time. However, they did not test the items individually. They found the best-fitting
model to be the Simms et al. (2002) model (Model 5 in this paper). Differing results
between the studies may be related to sample characteristics. Baschnagel studied
college students in Western New York 1- and 3-months after the terrorist attacks on
24

September 11
th
, 2001, while the present study sampled treatment-seeking women with
co-occurring disorders who reported many different types of traumatic experiences
across their lifetimes.
The importance of establishing metric invariance across time should not be lost
on clinicians or researchers as simply a psychometric exercise. The value added by
assessing this stringent level of longitudinal consistency is that if it can be established,
changes in scale scores across time can be more easily interpreted as changes in the
underlying construct (i.e., posttraumatic stress symptoms) rather than to person specific
variation, measurement error or a combination of effects. At the item level, we are
saying conceptually that what a person experiences as an exaggerated startle response at
time 1 is the same experience at time 2. While metric invariance can be a difficult
measure to achieve for psychological construct, it should not be overlooked.
Although it is unclear why the five items identified performed inconsistently
across time, a number of points may be raised. First, changes in how items related to
their factor across time could be influenced by treatment changing some women’s
understanding of how symptoms related to one another. Women in the current study all
began in the study in a residential treatment program, while some women also received
additional treatment focused on understanding trauma, PTSD, and substance abuse,
specifically. The PSS-SR assessed how bothered people are by each symptom. Based
on experiences in treatment, some women may have developed greater understanding of
how some of the posttraumatic stress symptoms related to their past traumas and current
25

psychological functioning, and, thus, reported them as less bothersome but not to all
symptoms in the cluster. Also, the items with factor loading that varied across time
most often referred to traumatic events specifically. Participation in trauma-informed
treatment may have influenced women’s understanding of items more closely tied with
traumatic materials compared to other items measuring broader construct such a diffuse
physiological arousal or emotional numbing.
Two psychometric concerns also may have influenced item functioning.. First,
four of these five items have compound stems, e.g., item 7, “trying to avoid activities,
people or places that remind you of the traumatic events”. While these are not the only
the compound items on the scale, problematic item performance may be related to items
containing more than one component. For example, on this item, at one time of
measurement, a woman might be responding with respect to people associated with her
traumatic events that she is trying to avoid while at another time she might be thinking
about activities she may be avoiding.
A second psychometric issue is the need to have an adequate number of items
for each construct. The clearest indication of this was for Effortful Avoidance, which
only had two items. A sufficient number of items is needed to assure that we are
accurately assessing the latent trait we are seeking to model with our manifest variables.
This is especially important as avoidance is believed to be a key mechanism in
maintenance of PTSD (Foa, Huppert, & Cahill, 2006; Foa & Kozak, 1986).
26

Overall, these results and the poor fit of the one factor model suggest that
posttraumatic stress symptoms, as currently measured, do not represent a unified
construct. Much of the recent research on the factor structure underlying posttraumatic
stress symptoms and PTSD has pointed towards a four-factor structure. The four-factor
structure suggests that there are multiple important underlying processes in the
expression of PTSD. Because of the inconsistency in the items making up the Effortful
Avoidance subscale, we could not fully investigate the adequacy of a four factor
solution longitudinally in this sample. However, the two-factor solution identified as
our final model was similar to the alternating symptom patterns described by Horowitz
(1986) in his psychodynamic model of PTSD. The two-factor solution may also be
considered in line with cognitive-behavioral understandings of PTSD. The Re-
experiencing/Avoidance factor is reflective of Emotional Processing Theory for PTSD
(Foa et al., 2006; Foa & Kozak, 1986) with the development of a fear structure
following trauma and the pathological avoidance of triggering it. The
Numbing/Arousal factor may represent the social and cognitive results of challenges to
personal belief systems following traumatic exposure and the associated negative affect
(Brewin, Dalgleish, & Joseph, 1996; Rothbaum, Meadows, Resick, & Foy, 2000). One
important clinical implication is that various underlying PTSD processes may be
differentially affected by different treatment components.
There are several limitations in this study. First, the sample consisted of only
women and, specifically, women who had a substance abuse disorder, an additional
27

mental health disorder and histories of numerous types of physical, emotional and
sexual abuse. How these women reacted to the posttraumatic stress scale may be
influenced by the level and breadth of the challenges they face. However, as there are
4.2 million adults in the U.S. with co-occurring mental health and substance abuse
disorders (Substance Abuse and Mental Health Services Administration, 2007), this
remains an important population. Also, this sample included only women whose
baseline assessment coincided with starting a treatment program. Consequently, the
sample may not represent the population of traumatized individuals as a whole, and
participation in treatment could have affected the psychometric outcomes. Also, one
could argue that the PSS-SR is only meant as a screening measure for PTSD and that
metric invariance should not be expected. However, the PSS-SR remains a commonly
used scale for both screening and for assessing outcomes; therefore, consistent item
performance across time is a worthy goal. Finally, dropping items with factor loadings
that varied across time is not the only, or necessarily best, way of dealing with time-
varying items. Variation in these items may be related to either specific factors
associated with items or with error variance – and these two sources of variability
cannot be differentiated in the current models. In addition, because the common factors
identified are partly determined by all of the items included in the scale, dropping items,
or ignoring the unique variances, is not recommended and including the specific factors
in the model can be theoretical meaningful (Meredith & Horn, 2001).
28

Despite the limitations, this study has several advantages. It is the first study to
look at the psychometric functioning of the PSS-SR across more than two time points.
It replicates the earlier findings by Baschnagel et al. (2005) that people do not respond
to some of the PSS-SR items consistently across time. It also extends this work by
identifying specific items that are not performing consistently, which suggests areas of
future scale development. Finally, the population studied is a previously unstudied,
large and relevant group.
In summary, we investigated the longitudinal consistency and the underlying
factor structure of the PSS-SR in a group of 370 women with co-occurring disorder and
histories of interpersonal violence. Our results indicate that even if a scale has
repeatedly been shown to be reliable at any single time point, it does not guarantee that
the scale is reliable across multiple time points. The final 12 item scale comprised of
two factors (Re-experiencing/ Effortful Avoidance and Numbing/Arousal) performs
consistently across time and can be used to evaluate change in response to treatment.

29

Chapter 3: Impact of trauma-focused treatment on trauma-related
outcomes among women with co-occurring disorders and histories of
interpersonal violence

Chapter 3: Abstract
Until recently, many researchers and treatment providers discouraged simultaneous
treatment of both mental health and substance abuse disorders. This study investigated
whether attendance in a trauma-focused treatment designed for individuals with
comorbid disorders (Seeking Safety) and level of attendance predicted changes in PTSD
symptoms and unsafe events among 370 women in residential treatment for comorbid
disorders and histories of interpersonal violence. Results indicated that there was a
significant drop in PTSD symptoms across time, however when statistically controlling
for relevant variables, extent of attendance in trauma-focused treatment did not
significantly predict changes in symptoms. There was a significant reduction in unsafe
events at six-months, with more trauma-focused treatment predicting fewer unsafe
events, and these gains were maintained at12-months.

30
Chapter 3: Introduction
Interpersonal violence is a widespread phenomenon among women and girls in
the United States. The rate of being a victim of physical and/or sexual assault among
females, as a child or an adult, is 13% - 40% (Coker, Smith, McKeown, & King, 2000;
Tjaden & Thoennes, 2000), However, this is likely an underestimation due to
underreporting (Commonwealth Fund, 1996). The multi-faceted impact of
interpersonal abuse, especially childhood abuse, has been widely documented (Becker
et al., 2005; Briere, Woo, McRae, Foltz, & Sitzman, 1997; Clay, Olsheski, & Clay,
2000; Cloitre, Cohen, Edelman, & Han, 2001; Courtois, 1988; Harris & Fallot, 2001;
Wilsnack, Vogeltranx, Klassen, & Harris, 1997), including physical, medical,
psychological, educational, occupational and relationship problems (Cunningham,
Pearce, & Pearce, 1988; Golding, 1999; Larson et al., 2005; Saunders, Villeponteaux,
Lipovsky, Kilpatrick, & Veronen, 1992; Stein & Kennedy, 2001; Terr, 1991; van der
Kolk, 1996a). Within this population, women also often suffer comorbid mental health
challenges, especially “co-occurring disorders” or the combination of a substance abuse
disorder and another mental health disorder, such as Posttraumatic Stress Disorder
(PTSD) and other anxiety disorders, Major Depressive Disorder and other mood
disorders, and other types of psychopathology (Hendrickson et al., 2004). Recently,
developing integrated treatment programs for women with co-occurring disorders has
become an important area of research and treatment (Hendrickson et al., 2004). It was
within the context of the assessment of one such integrated treatment program that the

31
current study was conducted. Specifically, this study investigates whether participation
in a trauma-focused group treatment as well as one’s level of participation in this group
predicted significant changes in psychosocial functioning above and beyond any gains
made by women who were participating in the larger integrated treatment program. Our
assessment focused on the effect of trauma-focused treatment on two trauma-related
outcomes: did a treatment protocol focusing on both trauma and substance abuse help
women with co-occurring disorders: 1) reduce posttraumatic stress symptoms during
treatment and were any gains maintained across time and 2) maintain greater levels of
personal safety by reducing the frequency of stressful and traumatic events, both during
treatment and in later months.
Impact of co-occurring disorders on women
The number of women struggling with co-occurring mental health and substance
use disorders as well as with histories of trauma is substantial (approximately 10
million, U.S. Department of Health and Human Services, 1999). In addition, there is
evidence that the occurrence of any of these challenges alone puts a woman at higher
risk for additional negative outcomes, such as subsequent traumas, increased likelihood
of substance use in response to trauma or mental health concerns, and development of
additional mental health disorders (Coid et al., 2001; Coker et al., 2000; El-Bassel,
Gilbert, Wu, Go, & Hill, 2005; Kilpatrick, Acierno, Resnick, Saunders, & Best, 1997);
thus, creating an increasingly larger set of challenges for women with already limited
coping resources. Research has shown that women with co-occurring disorders and

32
histories of trauma report more severe difficulties and use more services compared to
women with any one of these problems alone (P. J. Brown et al., 1995; V. B. Brown,
Huba, & Melchior, 1995; Harris, 1994; Timko & Moos, 2002). In addition, this group
generally has an extremely high treatment drop out rate (Comfort & Kaltenbach, 2000).
Psychological treatment for co-occurring disorders
Unfortunately, current services and programs designed to help women with co-
occurring disorders and histories of trauma are often inadequate in meeting the core
needs of these women (Greenfield et al., 2007; Grella, 1996; Harris, 1994; Harris &
Fallot, 2001; Holdcraft & Comtois, 2002; N. K. Young & Grella, 1998). Many of the
services currently available focus only on the immediate physical safety needs of
survivors, do not screen for a history of violence in mental health or substance abuse
treatment or fail to take into account the effects of previous trauma on treatment (Elliott
et al., 2005; Harris & Fallot, 2001; Zweben, 1996). This lack of attention to women’s
trauma history and trauma-related symptoms can interfere with a woman’s willingness
to participate in mental health or substance abuse services as well as increasing the
likelihood of relapse (P. J. Brown, 2000; P. J. Brown et al., 1995; Melchior, Huba,
Brown, & Slaughter, 1999). In addition, most services for mental health, substance
abuse, and trauma are provided by different service systems with different eligibility
criteria and procedures, and little coordination among service agencies (Ridgely,
Goldman, & Willenbring, 1990; Salasin, 2005; Watkins, Burnam, Kung, & Paddock,
2001; N. K. Young & Grella, 1998).

33
In response to an increasing awareness of the negative impact of co-occurring
disorders in the lives of women and their children, there has been a move towards
developing and testing integrated service models (Alexander, 1996; RachBeisel, Scott,
& Dixon, 1999; Watkins et al., 2001). Congressional and executive orders starting in
the 1990’s charged the Substance Abuse and Mental Health Services Administration
(SAMHSA) with improving programs for the treatment of substance abuse and mental
health as well as focusing on the needs of women (Salasin, 2005). Women consumers
of services lobbied providers, administrators, and researchers to design comprehensive
programs that included trauma treatment along with other mental health and substance
abuse services. Within this context, SAMHSA initiated the Women, Comorbid
Disorders and Violence Study (WCDVS, Giard et al., 2005; McHugo, Kammerer et al.,
2005) to develop and investigate new models of integrated treatment. Program
development was guided by four principles; 1) organizations and services must be
integrated to meet the needs of this population; 2) settings and services must be trauma-
informed; 3) consumer/survivor/recovering persons must be integrated into the design
and provision of services; and 4) a comprehensive array of services must be offered
(Huntington et al., 2005).
WCDVS was the first large scale federally funded study to investigate the
impact of integrated treatment programs for women with comorbid disorders and a
history of interpersonal trauma (McHugo, Kammerer et al., 2005). Each intervention
site was required to provide eight core services, including resource coordination and

34
crisis intervention, staff knowledgeable about trauma, holistic treatment of mental
health, trauma, and substance use issues as well as maintaining active involvement of
consumers in service planning and provision (Morrissey, Jackson et al., 2005).
However, because there are no definitive data on the best implementation of these
services, each site designed their own implementation of these core services (McHugo,
Kammerer et al., 2005). The study used a quasi-experimental design, in that it did not
include random assignment of women to treatment groups. Also, it used an intent-to-
treat design, which followed each woman across the 12 months of the study,
independent of whether she remained in treatment or not. Women were interviewed on
as many as five occasions; at baseline, and then at 3-months, 6-months, 9-months and
12-months following enrollment in the study. Measures taken at 3- and 9-months
focused on assessing services used by each woman. Interviews at 6- and 12-months
were more extensive, and included mental health, trauma and traumatic symptoms
measures as well as substance use measures, which were all in the baseline interview.
Across all nine national sites, small but significant reductions in mental health
symptoms and substance abuse were identified at 6-months at intervention sites in
comparison to treatment-as-usual sites (Cocozza et al., 2005; Morrissey, Ellis et al.,
2005). Results of prospective meta-analyses showed significant reductions in post-
traumatic stress symptoms at 6-months in sites that provided a higher level of integrated
services (Cocozza et al., 2005). Similar results were found at 12-months, with
significant reductions in overall mental health and posttraumatic stress symptoms found

35
for intervention sites, especially when the contrast for the level of integrated services
was the highest (Morrissey, Jackson et al., 2005). Women moved from significant
impairment related to posttraumatic stress symptoms to moderate levels of symptoms.
The national WCDVS also found that recent hospitalization for abuse, trauma or
violence negatively impacted the positive impact of integrated treatment programs on
posttraumatic stress symptoms (Morrissey, Jackson et al., 2005), though the authors
note that this result is based on a small number of women (3%) who reported recent
hospitalization due to trauma-related events. In studies of Los Angeles site data from
the WCDVS (Gatz et al., 2007; Gatz et al., 2004), women in the integrated treatment
group showed better retention during the first three months of the study, and lower
posttraumatic stress symptoms and increased coping skills use, both at 12-months.
Trauma-focused treatment
While we know there were small but significant gains made by women in the
integrated treatment groups both within the WCDVS as a whole (Cocozza et al., 2005;
Morrissey, Jackson et al., 2005) and at the Los Angeles site alone (Gatz et al., 2007;
Gatz et al., 2004), it remains unclear which aspects specifically contributed to the
reduction in posttraumatic stress and other mental health symptom. Specifically, in
addition to the array of integrated mental health and substance abuse services offered to
the women at the integrated treatment programs in Los Angeles, women also
participated in a trauma-focused treatment group designed for individuals with
substance abuse disorders as well as trauma and PTSD, called Seeking Safety (Najavits,

36
2002). Using post hoc analyses, the present study sought to isolate the effects of the
trauma-focused treatment (Seeking Safety) that was used at the Los Angeles site of the
WVCDS on changes in trauma-related variables (i.e., posttraumatic stress symptoms
and frequency of stressful and traumatic events). We focused on whether women’s
participation in Seeking Safety and how many sessions attended impacted trauma-
related outcomes, above and beyond any gains related to participation in the larger
WCDVS treatment groups and any other residential treatment programs. In this study,
the number of days spent in residential treatment was used as a proxy for other types of
mental health and substance abuse services women used.
Seeking Safety (Najavits, 2002) is a manualized treatment program that teaches
cognitive, behavioral, and interpersonal skills focused broadly on increasing women’s
skills to establish safety, including safety from substances, improving interpersonal
skills to help women disengage from unsafe relationships (in cases of both domestic
violence and with substance-abusing friends), and developing better emotion regulation
and coping skills to avoid extreme psychological and behavioral symptoms, such as
dissociation and self-harm (Najavits, 2007). Seeking Safety is a first-stage trauma
therapy, being present-oriented, and focusing on psychoeducation about trauma, PTSD
and substance abuse as well as the development of emotion regulation and coping skills
(Najavits, 2002, 2007). Deficiency in affect regulation and interpersonal skills among
women with histories of childhood trauma is well-documented (Cloitre, Scarvalone, &
Difede, 1997; Linehan, 1993) and may represent one difference related to the long-term

37
effects of childhood or adult onset trauma (Cloitre, Koenen, Cohen, & Han, 2002; Van
der Kolk, 1996b). Several treatment programs for individuals suffering from co-
occurring problems, or the severe psychopathology related to long-term trauma, focus
on helping the individual develop three types of safety skills (i.e., skills related to
maintaining better safety, increasing interpersonal competence, and enhancing
emotional regulation), prior to working directly with traumatic material (Bradley &
Follingstad, 2003; Cloitre et al., 2002; Linehan, 1993; Najavits, 2002; van der Hart,
Nijenhuis, & Steele, 2006). Focus on avoidance reduction through emotional
processing of trauma materials (i.e. second stage procedures) are not addressed in this
therapy; though a trauma processing component has been successfully been
incorporated into this treatment program (Najavits, Schmitz, Gotthardt, & Weiss, 2005).
The Seeking Safety protocol includes a detailed treatment manual with 24
separate sessions. Each session includes four components; 1) Check-in, which focuses
on reinforcing each woman’s use of positive coping skills during the previous week or
identification of better coping skills when poorer coping occurred; 2) Quotation, reading
and discussion of a quotation designed to emotionally engage the clients and possibly
provide a source of inspiration; 3) Session Content, reviewing session handouts,
connecting the content to current lives of the women in the group and identifying,
describing and practicing positive coping skills; and 4) Check-out, each women
identifies something learned during the session and creates an action plan (i.e.,
homework assignment) for implementing skill(s) learned. Seeking Safety is designed

38
for flexibility of implementation with no set topic order, independent session topics, and
ability to use across a variety of settings, including with both groups and individuals, in
outpatient or inpatient settings, and with mixed and gender-specific groups.
Four of the nine sites in the WCDVS modified the Seeking Safety protocol for
use in the Integrated Treatment Groups (V. B. Brown et al., 2007). At the Los Angeles
site, Seeking Safety was offered to women initialed enrolled in the Integrate Treatment
Group and was conducted as a group treatment, consisting of 31 sessions (see Appendix
A for description of sessions). Women attended 90-minute groups, which occurred
twice weekly, with attendance restricted to women enrolled in the WCDVS. Mental
health provider, substance abuse counselors, and consumer/recovery/survivor
consultants all provided recommendations concerning the selection of which sessions to
include, which sessions were split into two parts, and the session order, and was based
on the most prominent perceived needs of women at each facility (V. B. Brown et al.,
2007). The treatment program was highly rated by women in the WCDVS and women
described the important parts of the treatment as the “group experience” (e.g., feeling
safe and interacting with women who had similar experiences), learning new types of
coping skills, and learning about trauma, PTSD, and substance abuse (V. B. Brown et
al., 2007).
A number of studies have shown positive treatment outcomes using Seeking
Safety with various groups, including outpatient women (Cohen & Hien, 2006; Hien,
Cohen, Miele, Litt, & Capstick, 2004; Najavits, Weiss, Shaw, & Muenz, 1998), women

39
in residential treatment (V. B. Brown et al., 2007; Gatz et al., 2007), adolescent girls
(Najavits, Gallop, & Weiss, 2006), outpatient men (Najavits et al., 2005), male and
female veterans (Cook, Walser, Kane, Ruzek, & Woody, 2006; Desai, Harpaz-Rotem,
Najavits, & Rosenheck, 2008), incarcerated women (Zlotnick, Najavits, Rohsenow, &
Johnson, 2003) as well as successfully being combined with other treatment programs,
such as with Exposure treatment (Najavits et al., 2005) and Dialectical Behavior
Therapy skills training (Holdcraft & Comtois, 2002). These included two randomized
controlled trials (Hien et al., 2004; Najavits et al., 2006). Results from many of the
studies have shown consistent decreases in posttraumatic stress symptoms at post-
treatment time points (Cook et al., 2006; Desai et al., 2008; Hien et al., 2004; Najavits
et al., 2005; Najavits et al., 1998; M. S. Young et al., 2004; Zlotnick et al., 2003) and
earlier follow-up time points (Desai et al., 2008; Hien et al., 2004; Zlotnick et al., 2003).
Longitudinal courses of trauma-related variables
While the impact of trauma-focused treatment could be assessed across a great
number of different types of outcome measures, this study focuses on only two trauma-
related measures. First, consistent with much of the previous literature, a self-report
measure of posttraumatic stress symptoms was included. In addition, a more behavioral
measure was included, which consisted of selected stressful and traumatic events
representing a measure of women’s ability to change their overall level of safety. It
should be made clear that women are not being held responsible for additional traumatic
experiences. Rather, as the trauma-focused treatment program used is this study

40
focuses on increasing coping skills, helping women develop healthy interpersonal skills,
and reducing any harmful behaviors, it was hypothesized that greater practice of such
skills would decrease the likelihood of the number of different types of the selected
events. One of the strengths of the WCDVS was its longitudinal design. One important
issue in studying variables across time is developing an understanding of how variables
may systematically change across time and what factors strongly influence changes.
Thus, literature on the natural course of each outcome measure and the effects of
residential treatment on each variable are discussed as well as how the trauma variables
may interact with one another.
As the WCDVS used an intent-to-treat model, following women across the 12-
month study period independent of whether they remained in their original treatment
program or not, it is important to consider what would be the expected course of
posttraumatic stress symptoms. Like many aspects of this population, predictions about
the longitudinal course of posttraumatic stress symptoms during treatment are likely
complex. Research on the natural course of posttraumatic stress symptoms following a
single traumatic experience suggests that without further traumatic exposure,
posttraumatic stress symptoms are expected to lessen over time (Riggs, Rothbaum, &
Foa, 1995). However, among individuals who develop or have PTSD, high levels of
posttraumatic stress symptoms can be maintained for years and decades (Buckley,
Blanchard, & Hickling, 1996; McFarlane, 1988; Port, Engdahl, & Frazier, 2001).
While a number of theories for the development and maintenance of PTSD exist, more

41
and more focus on the role of avoidance as a key factor in maintaining posttraumatic
stress symptoms, whether is the avoidance of memories, thoughts, feelings, situations or
anything else related to the trauma (Foa & Meadows, 1997; Rothbaum et al., 2000; Van
der Kolk, 1996b). Thus, if women engaged in more avoidant acts, such as leaving
treatment early and returning to substance abuse, posttraumatic stress symptoms levels
would not be resolved, except for some possible effect of self-medication of substance
use on arousal symptoms.
Another important consideration is the higher rate of revictimization among
individuals who previously been traumatized (Classen, Palesh, & Aggarwal, 2005;
Gold, Milan, Mayall, & Johnson, 1994; Herman, 1992; Koenig, Doll, O'Leary, &
Pequegnat, 2004; Messman-Moore & Long, 2002), which would likely increase
women’s current levels of posttraumatic stress symptoms. Both the number of
traumatic experiences (Gold et al., 1994; Houskamp & Foy, 1991) and the number of
different types of traumatic events (Briere, Kaltman, & Green, 2008; Follette, Polusny,
Bechtle, & Naugle, 1996) have been shown to predict higher levels of posttraumatic
stress symptoms. In addition, the impact of stressful, but not necessarily traumatic
events, also may have an impact on those already traumatized (L. A. King, King,
Fairbank, Keane, & Adams, 1998; Mol et al., 2005; Yehuda et al., 1995).
Within residential treatment itself, a number of factors could influence
posttraumatic stress symptoms across time. Some studies have shown that longer
lengths of stay in residential treatment are predictive of better outcomes or completing

42
treatment (Gerstein et al., 1997; Simpson, Joe, & Rowan-Szal, 1997; Zhang, Friedmann,
& Gerstein, 2003), and posttraumatic stress symptoms specifically (Gatz et al., 2007;
Simpson et al., 1997). In addition, studies have shown posttraumatic stress symptoms
to increase for women engaging in substance abuse services, and it is hypothesized that
when self-medication effects of substance abuse are removed, then women tend to
experience more intrusive and hyperarousal symptoms (Jacobsen, Southwick, &
Kosten, 2001). Some studies have found that when first engaging in treatment related
to psychologically traumatic experiences, posttraumatic stress symptoms may initially
rise as information is recalled and emotional responses are triggered, because substance
withdrawal symptoms may be interpreted as anxiety or are experienced as emotional
triggers for traumatic memories (Jacobsen et al., 2001; Stewart, Pihl, Conrod, &
Dongier, 1998). However, symptoms generally decrease over time, perhaps due to
physiological rebound (Abueg & Fairbank, 1992). Poor coping and emotion regulation
skills are also important in the maintenance of PTSD across time, which is why these
are often targeted in first level trauma therapies. Thus, women engaging in trauma-
focused treatment that included skill development would be expected to have lower
posttraumatic stress symptoms across time.
Fewer predictions about the pattern of unsafe events across time can be made.
The rate of unsafe events would likely be higher in this population compared to the
general population, based on reports that the rate of revictimization among those
already traumatized is higher than among those who have never been traumatized

43
(Briere & Runtz, 1987; Gorcey, Santiago, & McCall-Perez, 1986; Van der Kolk,
1996b), However, the number of traumatic events would be expected to be low during
any 12-month period as well due to low base rates for many of these events. In
addition, time spent in controlled residential treatment settings would also likely lower
the frequency of unsafe events while women remained in controlled facilities compared
to less supervised housing arrangement. Nevertheless, a number of factors could lead to
the occurrence of several unsafe events. Higher levels of posttraumatic stress
symptoms may predict additional traumatic experiences, which may be due to risky
behaviors related to ongoing or a return to substance abuse (Stewart et al., 1998) or poor
coping skills can increase the likelihood of the women being in unsafe situations.
Finally, by engaging in treatment and seeking to cope with previous avoided problems,
women might recognize previously unaddressed issues, such legal concerns,
interpersonal challenges with family members and friends, low educational and
vocational attainment, medical concerns, and financial concerns.
Goal and hypotheses for this study
In an ideal study for assessing the effects of trauma-focused treatment on
trauma-related outcomes, women would be randomly assigned to treatment groups,
which includes one group focused on providing only trauma-focused therapy compared
to a comparison condition that was either a wait-list control group or included some
other activities not related to helping women develop coping skills or trauma therapy.
The only differences between treatment groups would be the therapeutic procedures and

44
skills taught that were included in the trauma-focused treatment program.
Randomization of women to treatment groups would help control for any possible
differences among women enrolled in the study by randomly distributing between
groups.
In regards to posttraumatic stress symptoms, among women in the trauma-
focused treatment group, after an initial increase in symptom levels often seen among
people first engaging in trauma-related treatment, symptom levels would be expected to
gradually decrease over time as women developed and successfully applied new coping
skills as they progressed through the trauma-focused treatment program and later in
their recovery; though weekly symptom reports may show sudden drops in symptoms as
skills were gained and emotional and cognitive reflections on successes were made. In
contrast, little systematic change in symptom levels over time would be expected in the
comparison condition. Statistically speaking, in this model, variables representing both
the treatment group (whether women received trauma-focused treatment or not) as well
as the number of trauma-focused sessions attended (level of participation) would be
expected to predict significant changes in posttraumatic stress symptoms, especially in
later time periods when differences in symptom levels would be expected to be greater.
Similarly, the frequency of different types of unsafe events would also be
expected to gradually lessen across time as a woman’s range of safe coping skills
improved and reliance on less safe coping methods, such as substance abuse and risky
behaviors associated with substance use, decreased over time. The differences in

45
frequency of unsafe events across time also should be predicted by both treatment group
and level of participation in the trauma-focused treatment group.
The current study assesses the effects of trauma-focused treatment, but does so
within the framework of the WCDVS. Our goal was to investigate the impact of a
woman’s participation in trauma-focused treatment (operationalized as participation in
Seeking Safety for women in the Integrated Treatment Group) on changes in
posttraumatic stress symptoms and on the frequency of unsafe events reported by
women over the 12-month study period. However, the trauma-focused treatment was
only one of the treatment component offered to women in the Integrated Treatment
Group of the larger WCDVS. Women in Integrated Treatment Group received a range
of services during their residential treatment, which could potentially impact both
posttraumatic stress symptoms and unsafe events. , Women in the Treatment-as-Usual
group did not receive Seeking Safety, but may have participated in a small number of
domestic violence classes. Thus, unlike the ideal study described above, the current
study included the number of days in women spent in any residential treatment settings
as a proxy variable for the broader treatment services that women received across the
course of the study. In addition, due to a number of factors, such as the quasi-
experimental design, large design of the study, women’s self-selected attrition, and the
complex nature of this population, clear and precise measures of the impact of the
trauma-focused treatment were difficult to separate out. Therefore, this study employed
a number of statistical procedures to identify, in so far as possible, the impact of

46
trauma-focused treatment on the two trauma-related outcome measures, including 1)
use of multivariate longitudinal structural equation models; 2) statistical adjustment for
time spent in residential treatment programs; and 3) statistical adjustment for WCDVS
treatment group.

47
Chapter 3: Method
Participants
This study used data from the Los Angeles cohort of the WCDVS (Gatz et al.,
2004). Women were interviewed an average of 18 days after entering both the
intervention (Integrated Treatment Group) and comparison (Treatment-as-Usual Group)
residential treatment agencies with 95% of the baseline interviews occurring within the
first month. The current study used data from 370 women who completed the baseline
interview and were in a residential treatment facility. A small number of women (n =
32) who were in the Los Angeles site dataset were not included in this study as they
were receiving outpatient services. (See Figure 3.1 for enrollment, study retention and
rates of trauma-focused treatment.) The average age in the sample was 33.2 years (SD
= 8.7). The sample was ethnically diverse; 25.7% Hispanic, 35.9% Caucasian, 23.2%
Black, 0.8% Asian or Pacific Islander, and 12.2% American Indian and 2.2%
multiracial or listed race as “other”. The mean education level was completion of 11.6
years of schooling (SD = 2.2). On a measure of stressful and traumatic life events,
women reported an average of 16 different types of events during their lifetime prior to
beginning the study. In addition, 73.9% reported some form of childhood abuse.
Prior analyses of the Los Angeles data of the WCDVS found few differences
between treatment groups of many variables (Gatz et al., 2004), including on the
following: age, education, race/ethnicity, relationship status (i.e., never partnered,

48
Figure 3.1. Outline of study enrollment, retention, and trauma-focused treatment

previously partnered or currently partnered), whether they have children or the number
of children, history of homelessness, presence of serious physical illness, measures of
general psychological functioning or psychiatric caseness (as measured by subscales of
the Brief Symptom Inventory, Derogatis, 1993), history of childhood abuse, level of

49
posttraumatic stress symptoms or baseline measures of coping skills. However, women
in the Integrated Treatment Group were more likely to be mandated to treatment, have
lower alcohol abuse scores (as measured by the Alcohol Severity Index, McLellan et al.,
1992) and were more likely to report methamphetamine and marijuana as their most
troublesome substance while women in the Treatment-as-Usual Group reported greater
use and difficulty with alcohol. Incarceration measures also differed between treatment
groups at baseline with women in the Integrated Treatment Group reporting that they
more likely to have ever been in jail and more likely to have recently been in jail
(within the last six months) compared to women in the Treatment-as-Usual Group (Gatz
et al., 2007).
It is unclear how the differential rates of mandated treatment, lifetime
incarceration and recent incarceration between Treatment Groups may affect treatment
outcomes. Supplemental analyses of incarceration for women included in this study
found the same differences in lifetime and recent (within the last six months)
incarceration rates. Further analyses showed that women in the Integrated Treatment
Group reported spending significantly more days in jail in the three months prior to the
study compared to women in the Treatment-as-Usual Group. Women in the Integrated
Treatment Group reported being incarcerated an average of 43.7 days (SD = 32.3, n =
98) in the three months prior to the beginning of the study and women in the Treatment-
as-Usual Group reported being in jail an average of 30.3 days (SD = 27.6, n = 64). In a
study of treatment retention of WCDVS data from Los Angeles and Boston sites,

50
lifetime and recent incarceration rates did not predict time in treatment, while women in
Integrated Treatment Group stayed in treatment significantly longer than women in
Treatment-as-Usual Group (Amaro, Chernoff, Brown, Arévalo, & Gatz, 2007).
However, because of the already highly complex nature of the models, no additional
variables, including those related to incarceration, were added to the current models to
adjust for these few baseline differences between groups.
Of the original Los Angeles site sample, 84% were interviewed at 12-months
(Gatz et al., 2007; Gatz et al., 2004). There was no difference in study retention rates
based on treatment group at 6-months (Treatment-as-Usual = 77% vs. Integrated
Treatment Group = 72%), but the retention rate in the study was significantly better for
the Treatment-as-Usual Group at 12-months (88%) compared to the Integrated
Treatment Group (79%). There were differences in the trajectory of treatment dropout
between conditions at 12-weeks into the study with a significantly lower drop out rate in
the Integrated Treatment Group (34.0%) compared to the Treatment-as-Usual Group
(41.5%, Gatz et al., 2007). Six months into the study, 29% of the women remained in
their original treatment program, while less than 10% were in their original treatment
program at 12-months. Roughly one in three women (34.5%) completed the treatment
program she began at the start of the study.
Because of the intent-to-treat design, differential attrition was investigated and
reported for the Los Angeles site data (Gatz et al., 2007; Gatz et al., 2004). Results
indicated a Time by Treatment Group interaction for posttraumatic stress symptoms,

51
indicating that in the Treatment-as-Usual Group, women who dropped out of treatment
reported significantly higher baseline symptoms compared to women who stayed in
treatment, while no effect was found for the Integrated Treatment Group. Women in
the Treatment-as-Usual Group reported significantly higher levels of both alcohol and
drug problems at baseline compared to the Integrated Treatment Group and, regardless
of condition, women who reported higher levels of drug problems at baseline were less
likely to remain in treatment. No differential attrition effects were identified for general
psychiatric functioning, psychiatric caseness or coping skills.
Measures
PTSD Symptom Scale – Self Report (PSS-SR). The PSS-SR (Foa et al., 1993) is
a 17-item measure consisting of the 17 cardinal symptoms included in the DSM-III-R
definition of posttraumatic stress disorder. Each symptom is rated on a 4-point scale
based on how bothered a person was by that the symptom during the last 30 days.
Anchors were not at all/only one time to 5 or more times per week/almost always. See
Appendix B for items. A refined measure of posttraumatic stress symptoms was used
as the measure of posttraumatic stress symptoms in this study (see Chapter 2 for a full
description of scale’s psychometric functioning). Five of the 17 original items were
dropped in order to maintain a consistent longitudinal measure; four items from the Re-
experiencing/Effortful Avoidance factor (items 2, 3, 5, and 7) and one item from the
Avoidance/Numbing factor (item 17).

52
Unsafe events. Each woman’s history for stressful and traumatic events was
assessed via a modified version of the Life Stressor Checklist-Revised (LSC-R,
McHugo, Caspi et al., 2005; Wolfe & Kimerling, 1997). This measure assessed
whether 30 specific stressful and traumatic events had occurred during a woman’s
lifetime and in the last six months as well as responses to one open-ended item for any
other stressful event. A shortened version of the LSC-R was included in the 6- and 12-
month interviews, which contained only items assessing if each event had occurred in
the previous six months. It is important to note that responses are dichotomous in
nature (i.e., did the event happen or not) and did not assess the frequency of the stressful
or traumatic event. Fifteen of the stressful and traumatic experiences from the LSC-R
were used to measure retraumatization and stressful events during each six month
period (see Appendix C for complete list of events included in scale). The events
included in this variable are those types targeted for reduction in the trauma-focused
treatment group. However, it needs to be clear that if these events did occur, the
occurrence was not always because of the direct actions of the women (which would
amount to “blaming the victim”), especially for retraumatization items, such as sexual
assault, physical abuse, and robbery. These types of events are outside the control of
individuals and blame for the event belongs to the perpetrator of the actions. However,
the trauma-focused treatment program was designed to enable individuals to reduce
unsafe behaviors that can put someone at greater risk for traumatic and stressful events,
and, thus, a decrease in the number different types of events across could be expected.

53
This variable is called unsafe events and was scored as the number of different types of
stressful and traumatic events reported during each six-month period; if the event
occurred it was scored as “1” and if it did not occur it was scored as “0”
1
. Thus, if
treatment had positive effects on unsafe events, the number of events would be expected
to be reduced.
Residential treatment. The measure of residential treatment (for either mental
health or substance abuse problems) was the sum of the number of days a woman
reported receiving residential treatment services during each three month period. These
data are derived from the Service Utilization section of the WCDVS interview that was
included at all five time points (Chung, Domino, Jackson, & Morrissey, 2007). Women
were asked if they had “received any residential treatment services,” and if so, where
and for how many days was the woman at each facility. When a woman reported being
at a residential facility and either did not know or refused to answer concerning the
number of days at the facility, the minimum number of days reported in any residential
facility for the sample was entered, one day. The number of days each woman spent in
residential treatment was included as a control variable, as women were enrolled in
residential treatment programs as part of the WCDVS. Also, more days in residential
treatment would likely affect both outcome measures, posttraumatic stress symptoms

1
The original study design included another measure of stressful and traumatic events consisting of the
remaining items of the LSC-R, and was considered a control measure as it may impact posttraumatic
stress symptoms. However, subsequent analyses indicated that the variable did not add significantly to
the model, and, thus, was dropped to simplify modeling and presentation of results.

54
and unsafe events
2
. It is important to note that women could have attended multiple
residential programs over the course of the study, and the data used here represent total
days spent in any program, not necessarily one continuous stay at their original
program. Also, the number of days women spent in other controlled environments,
such as jail or inpatient hospitalization, is not included in this variable. Finally,
treatment programs differed in length by their own design with the Treatment-as-Usual
Group tended to be shorter in length (6 to 9 months) compared to Integrated Treatment
Group (12 months). However, by design, the Treatment-as-Usual Group also tended to
discharge women to controlled settings, which would be included as residential
treatment in this study; thus, possibly mitigating differences based on differences in
treatment length of women’s original programs.
Treatment group. A dichotomous variable representing which WCDVS
treatment group a woman belonged to was included in later analyses. This variable
indicated whether the woman was in the Treatment-as-Usual Group (comparison) group
or was in the Integrated Treatment Group (intervention group, which combined mental
health and substance abuse services). Only women in the Integrated Treatment Group
were offered trauma-focused treatment (i.e., Seeking Safety classes) as part of the study
design. It is important to note that women were not randomly distributed into treatment
conditions, but were recruited from within the treatment program they had chosen to
attend.

2
The original study design included a second control measure of treatment that estimated amount of
treatment in uncontrolled settings. Based on analyses, the variable did not add significantly to the
modeling results, and, thus, was dropped to simplify the analyses and presentation of results.

55
Trauma-focused treatment. Trauma-focused treatment was measured as each
woman’s attendance in the Seeking Safety treatment program (Najavits, 2002), which
was offered to women in the Integrated Treatment Group (V. B. Brown et al., 2007).
For hypothesis testing, the number of Seeking Safety sessions a woman attended during
1) the first three months of the study and 2) during the second three months of the study
was used to assess the effects of trauma-focused treatment on outcome measures
(subsequently referred to as “Seeking Safety attendance”). Data were divided into
three-month intervals to mirror the timing of data collection of posttraumatic stress
symptoms. Nearly all of the women had completed their Seeking Safety classes by the
end of the first six months of the study. These data were collected during chart review
of each woman’s medical file. Overall, women in the Integrated Treatment Group
averaged 16.6 classes (SD = 12.4). The average number of classes during the first time
period was 10.1 classes (SD = 7.8) and during the second time period it was 6.0 classes
(SD = 6.9). The range of total number of Seeking Safety classes taken was 0 to 43.
While there were only 31 sessions in the series, a small number of women retook part of
the series.
Previous studies have investigated the effects of completing a certain number of
Seeking Safety sessions as a minimum dose. Najavits and colleagues (1998) used a
minimum of six classes while the WCDVS used 16 classes (or half of the available
sessions). Women in the current study who completed the minimum dose of six
sessions attended an average of 22.5 sessions, while women who completed a minimum

56
of 16 sessions attended an average of 27.6 sessions.
3
Of the 155 women enrolled in the
Integrated Treatment Group, 72.3% completed at least six sessions and 51.0%
completed at least 16 session of Seeking Safety.
Data Analyses
Assessment of the impact of trauma-focused treatment of posttraumatic stress
symptoms and occurrence of unsafe events proceeded in a step-wise fashion of building
increasingly more complex models based on the results of less complex models. Latent
factor structural equation modeling was done using Mplus 5.0 (Muthén & Muthén,
2007).
Univariate latent growth curve models were created for each of the two main
outcome variables (i.e., posttraumatic stress symptoms and unsafe events) and the
longitudinal control variable (days in residential treatment). Initially, this consisted of
creating longitudinal growth curve models using latent difference scores (LDS), which
were used to assess systematic intra-individual and inter-individual change in variables
across time (L. A. King et al., 2006; McArdle & Hamagami, 2001). In the LDS models,
observed scores are first partitioned into two components: the true score and
measurement error:
Y
0
= y
0
+e
0

and
Y
1
= y
1
+e
1

3
Analyses of the data based on completing Seeking Safety at minimum doses of 6, 16, and 24 sessions as
well as for total number of classes taken were conducted. Results did not differ from those presented in
this paper, and, therefore, are not presented. However, results are available upon request.

57
In these equations, y
0
and y
1
are true scores at the first and subsequent measurement
times, while e
0
and e
1
represent measurement errors for these same time periods. The
LDS for each longitudinal variable is then calculated as the difference between the true
scores, not including measurement error. For example, LDS for two time periods would
be calculated as:
LDSy
1
= y
1
– y
0

and
LDSy
2
= y
2
– y
1

In addition to external influences (e.g., Treatment Group, Seeking Safety
attendance, and other longitudinal variables), two internal sources of change on LDS for
each variable were investigated (McArdle & Hamagami, 2001). The first source of
change is the additive or constant change across time (e.g., “natural change”; L. King et
al., 2006) associated with the typical functioning of each variable across time. Previous
research has shown that posttraumatic stress symptoms tend to decrease across time
when no other traumatic events are present (Riggs et al., 1995). Thus, a negative slope
across time would be expected for posttraumatic stress symptoms in this study, if no
additional unsafe events occurred. In the longitudinal growth curve models, these types
of changes are modeled by predicting LDS via the estimated slope, represented as α
parameters in this study. Models with just this effect are called Constant Change
Models. Usually these effects are equated across time, but the equality constraints can
be relaxed to estimate non-linear growth patterns. The second type of internal source of

58
change in each longitudinal model are the autoregressive or proportional effects, which
assess the impact of one’s score in the preceding time point on one’s score at the current
time point. These models are referred to as Proportional Change Models. Proportional
effects were estimated as scores at the previous time point predicting LDS at the current
time point, and labeled with βs. A Dual Change Model included both Constant and
Proportional Change effects. The baseline comparison model, No Change Model,
contains only the estimate of the initial level and slope of each variable and estimates
neither the constant nor proportional changes on LDS. Figure 3.2 depicts the Dual
Change Model for posttraumatic stress symptoms. Basic model fit was assessed via a
number of standard model indices: each model’s chi square value ( χ
2
) and its associated
degrees of freedom, Akaike Information Criterion (AIC, Akaike, 1973), Bayesian
Information Criterion (BIC, Schwarz, 1978), Comparative Fit Index (CFI, Bentler,
1990a), Root Mean Square Error of Approximation (RMSEA, Steiger, 1990), and
Standardized Root Mean Square Residual (SRMR, Bentler, 1990b). With AIC and BIC
there is no predetermined value to identify better versus worse fitting models, however
smaller values are better. A CFI of at least .90 indicates adequate model fit (Bentler,
1990a). RMSEA values of .05 or smaller and SRMR less than .09 indicate good-fitting
models (Hu & Bentler, 1999).
Once univariate models were constructed, bivariate models of each combination
of the two main study variables (posttraumatic stress symptoms and unsafe events) and
the longitudinal control variable (days in residential treatment) were constructed.

59
Figure 3.2. Schematic of dual change model for posttraumatic stress symptoms

Notes. Squares = observed variables; Circles/Ellipses = latent variables; Triangle =
constant; One-headed arrows = regression coefficients; Two-headed arrows =
correlations, covariances or cross-products; f = fixed value; L
P
= latent posttraumatic
stress symptoms score; D
P
= latent difference score for posttraumatic stress symptoms;
subscript 1 – 5 = study time periods (1= baseline interview, 2 = 3-month interview, 3 =
6-month interview, 4 = 9-month interview, and 5 = 12-month interview); α =
coefficients representing constant or additive change; β = proportional or autoregressive
change.

Bivariate models included the cross-couplings or cross-lagged parameters representing
the predictive effects of one variable at the preceding time point on the other variable at
the current time point
4
. Next, an intermediary model with all three longitudinal
variables was constructed and all possible cross-couplings were tested. Once the best-
fitting longitudinal models were determined, next the effects of external variables (e.g.,

4
Results from stepwise model building for univariate, bivariate, and trivariate models are available upon
request.

60
Treatment Group and Seeking Safety attendance) were added to the univariate models.
Treatment Group and Seeking Safety attendance during the first three months of the
study were used to predict LDS at each time point. Seeking Safety attendance during
the second three-month period of the study was used to predict LDS for posttraumatic
stress symptoms and days in residential treatment for the last three LDS and both unsafe
events LDS. Final results for Treatment Group and Seeking Safety attendance on each
of the three univariate longitudinal variables are recapped below. Information gained
from analyses of Treatment Group and Seeking Safety attendance on the univariate
models were added to the three-variable longitudinal model and, where necessary,
additional parameters were added in order to appropriately test the study hypotheses.

61
Chapter 3: Results
Results are presented in a step-wise fashion. First, analyses assessing possible
effects of non-randomization of women into study conditions are presented. Second,
descriptive statistics and individual LDS models for each of the three longitudinal
variables (posttraumatic stress symptoms, unsafe events, and days in residential
treatment) are described. The effects of Seeking Safety attendance and Treatment
Group were added to the univariate longitudinal models, and final results are presented
below. The three univariate models were combined and appropriate cross-couplings
were added to allow for testing the study hypotheses, which are presented below.
Effects of Non-randomization of Women to WCDVS Treatment Groups
Table 3.1 summarizes differences in the three main study variables based on
Treatment Group. Women in the Treatment-as-Usual Group (n = 215) did not
significantly differ from women in the Integrated Treatment Group (n = 155) at baseline
on either the outcome measures (current posttraumatic stress symptoms and unsafe
events in the six months prior to the beginning of the study) or on the number of days
women spent in residential treatment settings during the first three months of the study.
The groups did differ significantly on the number of days women were in residential
treatment in the three months prior to the beginning of the study, with women in the
Treatment-as-Usual Group averaging about eight more days in residential treatment
compared to women in the Integrated Treatment Group. It is unclear how this
difference may affect outcomes on main study variables, but subsequent analyses

62
statistically adjusted for days in residential treatment settings in the three months prior
to the beginning of the study. In longitudinal models, this effect was modeled by
allowing the initial level of days in residential treatment to correlate with the Treatment
Group. The impact of this effect would likely be hard to disentangle from the different
lengths of treatment episodes already built into the program length. The differences
between groups in days spent in residential treatment services in the three months prior
to the beginning of the study may be offset by the differences between Treatment
Groups for incarceration in the three months prior to the study.

Table 3.1. Means (and SD) for Study Variables by Treatment Group.

Treatment-As-Usual
Group
N = 215

Integrated Treatment
Group
N = 155
Modified PSS-SR at BL

15.3 (8.8)

15.6 (7.5)

Unsafe Events Scale at BL
(six months prior to study)

3.8 (2.5) 3.6 (1.4)
Days in residential treatment
90 days prior to start of study

First 90 days of study

26 (20)*

75 (25)

18 (15)*

76 (24)

Notes. * signifies significant differences between groups based on t-tests, p < .05; PSS-
SR = Posttraumatic Symptom Scale – Self Report, BL = baseline interview.

63
Descriptive Statistics
Posttraumatic stress symptoms. Figure 3.3 depicts posttraumatic stress
symptoms for a random sample of 40 women from the entire sample, while Table 3.2
provides descriptive statistics for all longitudinal variables across the five time points
and by Treatment Groups. Inspection of the means suggests that after an initial
reduction in posttraumatic stress symptoms during the first three months of the study,
symptoms decreased only slightly across the remaining time periods with a slight
increase during the last six months. However, the individual data indicate a great deal
of variability in how scores changed within individuals across time. For some women
scores remained relatively stable, while for others their symptoms varied significantly.
This suggests the likely presence of several subpopulations in the data set; though it
remains unclear which factors differentiate by these subgroups.
Unsafe events. Figure 3.4 depicts raw data for 40 random cases and descriptive
statistics can be found in Table 3.2 for number of different types of unsafe events
women reported experiencing during the previous six months. As the graph and the
table suggest, women generally reported more unsafe events in the six months prior to
the start of the study compared to either the first or second six-month period of the
study. There was some variability during the second half of the study with some
women showing an increase in unsafe events while others reported fewer such events.

64
Figure 3.3. Posttraumatic stress symptom scores across time for 40 random cases

Notes. BL = baseline interview; 3-mo = 3-month interview; 6-mo = 6-month interview;
9-month interview; 12-mo = 12-month interview.

Figure 3.4. Number of different types of unsafe events for 40 random cases

Notes. BL = baseline interview; 6-mo = 6-month interview; 12-mo = 12-month
interview (data for unsafe events were not collected at 3- and 9-months).

65
Table 3.2. Means (SD and sample sizes) for Study Variables by Interview Period and Treatment Group.

Variable Baseline
Mean (SD, n)

3-months
Mean (SD, n)

6-months
Mean (SD, n)

9-months
Mean (SD, n)

12-months
Mean (SD, n)

PSS: all participants
Treatment-as-Usual Group
Integrated Treatment Group

15.5 (8.3, 370)
15.3 (8.8, 215)
15.6 (7.5, 155)

11.5 (8.2, 270)
10.7 (8.5, 161)
12.7 (7.7, 109)
11.3 (8.7, 277)
10.3 (8.7, 166)
12.7 (8.7, 111)
10.6 (9.8, 243)
9.9 (8.7, 147)
11.7 (11.1, 96)
11.0 (9.1, 265)
11.4 (9.6, 164)
10.4 (8.1, 99)
UE: all participants
Treatment-as-Usual Group
Integrated Treatment Group

3.7 (2.4, 370)
3.8 (2.5, 215)
3.6 (2.2, 155)
n/a 1.5 (2.0, 277)
1.5 (2.1, 166)
1.4 (1.8, 111)
n/a 1.6 (1.9, 303)
1.7 (1.9, 186)
1.3 (1.9, 117)
RES: all participants
Treatment-as-Usual Group
Integrated Treatment Group

22.8 (18.1, 370)
26.3 (20.0, 215)
17.9 (14.6, 155)
75.5 (24.9, 326)
74.9 (25.2, 186)
76.4 (24.5, 140)
54.0 (38.0, 286)
48.5 (37.3, 170)
62.5 (37.7, 111)
34.1 (40.0, 263)
24.0 (34.4, 161)
50.0 (43.0, 102)
20.3 (33.6, 272)
13.1 (26.9, 169)
32.1 (40.0, 103)
Seeking Safety Attendance* n/a 10.1 (7.8, 111)

6.0 (6.9, 117) n/a n/a

Notes. PSS = posttraumatic stress symptoms; UE = unsafe events; RES = days in residential treatment in each three-
month period; * Seeking Safety was offered only to women in the Integrated Treatment Group and each woman’s
frequency of attendance was recorded in two three-month intervals (baseline through 3-months and 4-months through 6-
months).

66
Days in residential treatment. Figure 3.5 depicts raw data for 40 random cases
and Table 3.2 displays sample means and SD for the number of days in residential
treatment across the five study periods and by Treatment Group. Both data summaries
suggest that women generally were in residential treatment for more days during the
first three-month period of the study. This was followed by a steady decrease in the
average number of days in residential treatment for each subsequent three-month
period. However, the raw data also suggest there is a lot of variability in how long
women were in residential treatment. Over the course of the year, it appears that many
women stayed in treatment over a number of three-month time periods and then
discontinued residential services, but the length of treatment differed greatly among
women. However, the variable is the summation of the number of days each woman
reported spending in any residential program, and cannot necessarily be interpreted as
one continual residential treatment episode, but could represent sequential enrollment in
a number of different programs.
Longitudinal Univariate Growth Curve Models
Posttraumatic stress symptoms. Seven longitudinal growth curve models were
developed to assess the trajectories of posttraumatic stress symptoms across the five
time points. Table 3.3 summarizes the modeling results. Model 3c, a Constant Change
Model, was selected as the final model for posttraumatic stress symptoms and was used
in all subsequent analyses. Model 3c was selected as the final model despite Model 4
(Dual Change Model) fitting the data statistically better based on chi square values.

67
Figure 3.5. Days in residential treatment for a random sample of 40 cases

Notes. BL = baseline interview; 3-mo = 3-month interview; 6-mo = 6-month interview;
9-month interview; 12-mo = 12-month interview.

The BIC was 14 points better for Model 3c, reflecting the fewer parameters in this
model, while the AIC, which does not take into account the number of model
parameters, was just one point different. In Model 4, none of the alpha or betas
coefficients were significant and it was decided that the added complexity was not
conceptually beneficial and thus the well-fitting Model 3c was retained as the final
model. In this nonlinear model, both the two middle bases were equated while the last
alpha parameter was allowed to vary. Figure 3.6 depicts the results for the final model
for the whole sample. These results, as illustrated in the raw scores depicted in Figure
3.3, suggest that posttraumatic stress symptoms did not change in a strictly linear
fashion. The mean level was 15.4, indicating that at Baseline the average score for this

68
sample was 15.4 points on this modified version of the PSS-SR. The mean slope of -3.1
points suggests that were moderate changes in posttraumatic symptom scores (in
relation to the Baseline mean of 15.4 points) between time periods. The directions of
the changes are a function of the combined signs of the mean for slope and the alpha
coefficients associated with the slope. The negative slope and first alpha coefficient of
1 indicate that posttraumatic stress symptoms tended to go down more sharply during
the first three months of the study compared to the second and third time periods (alpha
coefficients = .3). There were no significant changes in posttraumatic stress symptoms
during the last three months of study (alpha coefficient = -.2, p > .05). The SD of the
level is 34.0 and the SD of the slope is 8.4, which indicate that there are moderate
amounts of variability in both where each woman began on the posttraumatic stress
symptom scale and in the average amount of change across time.
Next, the potential impacts of Treatment Group, Seeking Safety attendance
during the first three months of the study and Seeking Safety attendance during the
second three-month time period on changes in posttraumatic stress symptoms were
assessed. The final model fit the data well, χ
2
(10) = 14.2, p = 16, CFI = 1.00, AIC =
14664, BIC = 14797, RMSEA = .03 (.00, .07), P(RMSEA< .05) = .73, SRMR = .03.
Figure 3.7 depicts the final modeling results. Results of the univariate analyses showed
that neither Treatment Group nor the Seeking Safety variables predicted significant
changes in posttraumatic stress symptoms across time. For presentation purposes,
Table 3.4 displays average difference scores in posttraumatic stress symptoms across

69
Figure 3.6. Final univariate longitudinal model of posttraumatic stress symptoms

Notes. * indicates significant parameter estimate, p < .05; parameters that share a
common superscript were equated; P
X
= posttraumatic stress symptoms at times 1-5;
time 1 = baseline interview; time 2 = 3-month interview; time 3 = 6-month interview;
time 4 = 9-month interview; time 5 = 12-month interview; L
Px
= latent factors for
posttraumatic stress symptoms at times 1-5; D
Px
= latent differences scores at times 2 –
5.

each three-month period by Treatment Group, Interview Period and Seeking Safety
attendance. Seeking Safety attendance in the Integrated Treatment Group was divided
into three subgroups based on total number of sessions attended: 1) 0-5 classes, 2) 6-15
classes, and 3) 16 or more classes (as the Treatment-as-Usual Group did not receive
Seeking Safety treatment per program design, their data could not be disaggregated
according to this variable). Inspection of the data shows an average drop of three points
in posttraumatic stress symptoms during the first three months of the study in the

70
Table 3.3. Longitudinal Latent Difference Score Models of Posttraumatic Stress Symptoms.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
1. No Change 153.3 (17)
< .001

n/a n/a .71 9895 9907 .15 (.13, .17)
< .001
.16
2a. Proportional Change
equated betas

109.6 (16)
< .001
1 43.7 (1)
< .05
.80 9853 9869 .13 (.10, .15)
< .001
.16
2b. Proportional Change
all betas freed

83.9 (13)
< .001

1

2a

69.4 (4)
< .05

25.7 (3)
< .05
.85 9833 9861 .12 (.10, .15)
< .001
.16
3a. Constant Change
with all alphas freed

18.5 (11)
.07
1 134.8 (3)
< .05
.98 9772 9807 .04 (.00, .08)
.59
.05

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df = degrees
of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison model;
CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; RMSEA =
root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that RMSEA is
equal to or less than .05; SRMR = standardized root mean square residual.

71
Table 3.3. Continued.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
3b. Constant Change
with last 3 alphas
equated

24.5 (13)
.03
1

3a

128.8 (4)
< .05

6.0 (2)
< .05

.98 9774 9801 .05 (.02, .08)
.48
.07
3c. Constant Change
with two equated
alphas*

19.2 (12)
.08
1

3a

3b

134.1 (5)
< .05

0.7 (1)
> .05

5.3 (1)
< .05
.98 9770 9802 .04 (.00, .07)
.65
.05

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df =
degrees of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison
model; CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion;
RMSEA = root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that
RMSEA is equal to or less than .05; SRMR = standardized root mean square residual.

72
Table 3.3. Continued.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
4. Dual Change Model

9.3 (8)
.32

1

2b

3c

143 (9)
< .05

74.6 (5)
< .05

9.9 (4)
< .05
1.00 9769 9816 .02 (.00, .07)
.82
.03

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df =
degrees of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison
model; CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion;
RMSEA = root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that
RMSEA is equal to or less than .05; SRMR = standardized root mean square residual.
73

Treatment-as-Usual Group and two of the three Integrated Treatment Groups, those
with the highest and lowest levels of Seeking Safety. In general, it appears that all
groups tended to maintain any reductions in posttraumatic stress symptoms across time
with the Treatment-as-Usual Group showing a slight, but not significant, increase in
symptoms during the last three months of the study.

Figure 3.7. Depiction of partial modeling results for posttraumatic stress symptoms,
Treatment Group, and Seeking Safety attendance

Notes. * indicates significant parameter estimate, p < .05; D
Px
= latent difference score
for posttraumatic stress symptoms at times 2 - 5; time 2 = change in scores between
baseline and 3-month interviews; time 3 = change in scores between 3-month and 6-
month interviews; time 4 = change in scores between 6-month and 9-month interviews;
time 5 = changes in scores between 9-month and 12-month interviews.

74

Table 3.4. Means of Difference Scores for Posttraumatic Stress Symptoms by
Treatment Group, Interview Period, and Seeking Safety Classes
5
.

Treatment Group
Total Classes

PSS (3-mo –
BL)

PSS (6-mo –
3-mo)

PSS (9-mo – 6-
mo)

PSS (12-mo –
9-mo)

Treatment-as-Usual
Group
0 classes
Mean = -4.4
SD = 9.0
n = 161

Mean = -2.8
SD = 9.0
n = 161
Mean = -0.1
SD = 8.0
n = 147
Mean = 1.1
SD = 7.4
n = 138
Integrated Treatment
Group
0 - 5 classes
Mean = -2.5
SD = 7.3
n = 23

Mean = -3.1
SD = 10.4
n = 23
Mean = -1.2
SD = 8.9
n = 26
Mean = -1.2
SD = 6.8
n = 25
Integrated Treatment
Group
6 – 15 classes
Mean = 0.3
SD = 9.2
n = 16

Mean = -2.6
SD = 11.1
n = 16
Mean = -4.3
SD = 6.2
n = 11
Mean = 0.2
SD = 10.6
n = 9
Integrated Treatment
Group
16 or more
classes
Mean = -3.3
SD = 6.7
n = 70

Mean = -1.2
SD = 7.5
n = 70
Mean = -0.1
SD = 11.0
n = 59
Mean = -2.3
SD = 11.9
n = 56

Notes. PSS = posttraumatic stress symptoms; BL = baseline interview; 3-mo = 3-month
interview; 6-mo = 6-month interview; 9-mo = 9-month interview; 12-mo = 12-month
interview; Total classes = total number of Seeking Safety classes each woman in the
Integrated Treatment Group attended (women in the Treatment-as-Usual group were not
offered Seeking Safety).

Towards potentially differentiating the effects of Seeking Safety attendance
from time spent in residential treatment on posttraumatic stress symptoms across time,
additional descriptive analyses were conducted. Using the Seeking Safety attendance

5
When comparing mean differences between time periods from Table 2.2 to means of raw difference
scores as presented in Table 2.4, differences may arise from varying sample sizes in each type of
calculation. Difference scores require data be available from both time periods for each woman while
summary statistics in Table 2.2 include all available data at each time period. Longitudinal structural
equation models used all available data in estimation procedures.
75

groups for women in the Integrated Treatment Group described above as a model, a
similar variable was sought based on days in residential treatment, that could be used
across both Treatment Groups. No clear pattern for days in residential treatment
emerged using the three groups discussed above. However, using the WCDVS
definition of completion of Seeking Safety as attendance in 16 or more sessions
provided a useful way to divide groups. Women in the Integrated Treatment Group
who did not complete Seeking Safety averaged 90 days (SD = 62, median = 83, range =
5 – 182) in residential treatment facilities during the first six months of the study
compared to an average of 168 days (SD = 25, median = 182, range = 5 – 182) for
women in the Integrated Treatment Group who did complete Seeking Safety. A
minimum of 130 days of residential treatment during the first six months of the study
(baseline interview to 6-month interview) was selected as the (somewhat arbitrary)
division point between a “High Residential Service Use” Group and a “Low Residential
Service Use” Group. This division created groups similar, but not identical, to groups
based completion of Seeking Safety. Table 3.5 displays means of difference scores for
posttraumatic stress symptoms by Treatment Group, Interview Period, and Residential
Service Use Group. No clear pattern of differences is apparent. Women in the
Treatment-as-Usual Group who also were in the High Residential Service Use Group
showed large drops in posttraumatic stress symptoms during the first three months of
the study (average decrease of 5.4 points) compared to averages of two to three points
76

in the other three groups. No clear linear pattern in symptom reduction was identified
based on this viewing of the data.
Finally, additional structural equation models (not shown here) using only data
from the Integrated Treatment Group were developed to test whether Seeking Safety
attendance was associated with decreases in posttraumatic stress symptoms within the
Integrated Treatment Group only. Models indicated that neither of the Seeking Safety
attendance variables predicted significant changes in posttraumatic stress symptoms at
any time point. Seeking Safety attendance variables also did not predict the slope
coefficients for posttraumatic stress symptoms. The only finding relevant to the current
study was that higher levels of posttraumatic stress symptoms at the baseline interview
predicted significantly fewer Seeking Safety classes attended during the second three
month time period (parameter estimate = -0.3, p = .04); however the effect was not
significant for the effect of baseline posttraumatic stress symptoms levels on Seeking
Safety classes attended during the first three month time period (parameter estimate = -
0.2, p = .19).
Unsafe events. Six longitudinal growth curve models were tested to model the
trajectories of unsafe events across (see Table 3.6). The Constant Change Model was
selected as the final model for describing the longitudinal pattern in changes of unsafe
events. In this non-linear model, the first alpha coefficient was fixed to 1.0 and the
second was free to vary. These results, as illustrated in the raw scores depicted in
Figure 3.8, suggest that unsafe events did not develop in a strictly linear fashion. The

77
Table 3.5. Means of Difference Scores for Posttraumatic Stress Symptoms by Treatment Group, Interview Period, and
Use of Residence Services Group.

Treatment Group

PSS (3-mo – BL)

PSS (6-mo – 3-mo)

PSS (9-mo – 6-mo)

PSS (12-mo – 9-mo)

Treatment-as-Usual Group
Low Use of Residential
Services
Mean = -1.7
SD = 9.3
n = 58

Mean = -1.5
SD = 10.2
n = 58
Mean = 0.3
SD = 8.5
n = 65
Mean = 0.9
SD = 7.2
n = 59
Treatment-as-Usual Group
High Use of Residential
Services
Mean = -5.4
SD = 8.0
n = 84

Mean = -1.5
SD = 6.9
n = 84
Mean = -0.8
SD = 7.2
n = 72
Mean = 1.8
SD = 7.6
n = 69
Integrated Treatment Group
Low Use of Residential
Services
Mean = -2.0
SD = 8.2
n = 31

Mean = 0.3
SD = 9.5
n = 31
Mean = -2.6
SD = 6.0
n = 29
Mean = -1.7
SD = 7.2
n = 28
Integrated Treatment Group
High Use of Residential
Services
Mean = -3.0
SD = 6.7
n = 67

Mean = -0.7
SD = 6.8
n = 67
Mean = -0.2
SD = 11.0
n = 59
Mean = -1.9
SD = 12.0
n = 54

Notes. PSS = posttraumatic stress symptoms; BL = baseline interview; 3-mo = 3-month interview; 6-mo = 6-month
interview; 9-mo = 9-month interview; 12-mo = 12-month interview.

78
Figure 3.8. Final modeling results for unsafe events

Notes. * indicates a significant parameter estimate, p < .05; U
x
= unsafe events at times
1, 3, and 5; time 1 = baseline interview; time 3 = 6-month interview; time 5 = 12-month
interview (data were not collected at 3- and 9-months); L
Ux
= latent factors for unsafe
events at times 1, 3, and 5; D
Ux
= latent differences score at times 1, 3, and 5; paths with
no associated numeric were fixed to 1.0 (data on unsafe events were not collected at
times 2 and 4 by study design).

second alpha coefficient of near zero (-.01) signifies that, on average, there was no
systematic difference in the number of events women reported between the first six
months of the study compared to the second six months of the study, which can be seen
in the graphs of both means and raw data. Further inspection of the modeling results
provides some insight into both intra-individual and inter-individual reports of unsafe
events across time. First, the mean of intercepts (3.9) indicates that women were

79
reporting that they had experienced around four different types of stressful and
traumatic events in the six months before the study began. The value of -2.1 for the
mean slope suggests that women’s reports of unsafe experiences tended to decrease
sharply over time. The SD of the intercepts (3.9) and the SD for slopes (4.1) indicate
that there is fairly high amount of variability in both the number and magnitude of
change across time in the number of different types of unsafe events women reported
experiencing, compared to their mean values. The negative correlation between initial
level and mean slope (-3.0, p < .05) suggests that women who reported more different
types of unsafe events during the three months prior to the beginning of the study
reported larger reductions in unsafe events (i.e. slope) during the study period.
Next, models estimating the effects of Treatment Group and both Seeking Safety
attendance variables on unsafe events across time were tested. The final model for this
set of analyses appears in Figure 3.9. Inspection of the results indicated that more
Seeking Safety classes taken during each of three-month periods predicted significant
decreases in the number of different types of unsafe events reported during the first six
months of the study. However, neither of the Seeking Safety attendance variables
predicted changes in unsafe events during the second six months of the study. In
regards to Treatment Group, there was a significant negative effect of Treatment Group
on unsafe events during the first six months of the study with women in the Treatment-
as-Usual Group reporting a significant larger decrease in unsafe events during the first
six months of the study compared to women in the Integrated Treatment Group. There

80
Table 3.6. Longitudinal Latent Difference Score Models of Unsafe Events.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
1. No Change 303.7 (6)
< .001

n/a n/a .00 4331 4343 .37 (.33, .40)
< .001
.32
2a. Proportional Change
all betas equated

94.3 (5)
< .001
1

209.4 (1)
< .05
.05 4124 4140 .22 (.18, .26)
< .001
.18
2b Proportional Change
all betas freed

54.7 (4)
< .001
1

2a

249.0 (2)
< .05

39.6 (1)
< .05
.46 4086 4106 .19 (.14, .23)
< .001
.15
3a. Constant Change
with alphas equated
114.9 (3)
< .001
1

188.8 (3)
< .05
.00 4149 4172 .32 (.27, .37)
< .001
.18

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df =
degrees of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison
model; CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion;
RMSEA = root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that
RMSEA is equal to or less than .05; SRMR = standardized root mean square residual.

81
Table 3.6. Continued.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
3b. Constant Change
with alpha freed

1.5 (2)
.47
1

3a

302.2 (4)
< .05

113.4 (1)
< .05
1.00 4037 4065 .00 (.00, .10)
= .72
.03
4. Dual Change Model

0.1 (1)
.97

1

2b

3b

303.6 (5)
< .05

54.6 (3)
< .05

1.4 (1)
> .05
1.00 4037 4069 .00 (.00, .00)
.98
.01

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df =
degrees of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison
model; CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion;
RMSEA = root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that
RMSEA is equal to or less than .05; SRMR = standardized root mean square residual.

82

was an opposite trend at 12-months, in that women in the Integrated Treatment Group
reported larger reductions in unsafe events during the second half of the study compared
to women in the Treatment-as-Usual Group.
The combined effects of both Treatment Group and Seeking Safety attendance
on unsafe events during the first six months, as described above, appear contradictory,
but likely represent an interaction-like effect with Seeking Safety attendance accounting
for reductions in unsafe events among those in the Integrated Treatment Group, while
the similar level of reduction in unsafe events seen among women in the Treatment-as-
Usual Group was statistically ascribed to the Treatment Group variable. For
presentation purposes, Table 3.7 provides means of difference scores based on the raw
data for unsafe events by Interview Period, Treatment Group, and Seeking Safety
attendance. Inspection of the data for women in the Integrated Treatment Group
(presented in the second column of Table 3.7) suggests that a greater frequency of
attendance in Seeking Safety was related to a large decrease in the number of unsafe
events reported during the first six months of the study compared to women who
attended fewer Seeking Safety classes. Women who took the most Seeking Safety
classes reported a mean difference of -2.8 events compared to -1.0 events for women
with the fewest classes. This is consistent with the significant negative parameter
estimates from SC1 and SC2 to D
U3
. However, women in the Treatment-as-Usual
Group also reported large decreases in the number of unsafe events at 6-months
83

compared to baseline with an average of 2.0 fewer unsafe events. The means of the
difference scores for Treatment Groups are not different (Treatment-as-Usual Group =
-2.0, SD = 2.9 and Integrated Treatment Group = -2.1, SD = 2.5). Thus, while Seeking
Safety attendance accounted for reductions in unsafe events for women in the Integrated
Treatment Group, the significant parameter estimate linking Treatment Group with
unsafe events represents the significant reduction in unsafe events for women in the
Treatment-as-Usual Group not accounted for by any other explanatory variable in the
model. In summary, at 6-months women in both treatment groups reported decreases in
unsafe events compared to their baseline scores, and in the Integrated Treatment Group
this effect appears to be moderated by a beneficial effect of attending more Seeking
Safety classes.
Inspection of the average difference scores for unsafe events at 12-months
supported the trend seen in the modeling results with women in the Integrated
Treatment Group reporting somewhat larger reductions in unsafe events during the
second half of the study compared to women in the Treatment-as-Usual Group.
Inspection of the means of raw difference scores suggests there may be a non-linear
effect of Seeking Safety attendance at 12-months. Women with the largest reduction in
unsafe events at 6-months (i.e., those with the highest Seeking Safety attendance)
appeared to maintain this gain with a mean of difference scores of -0.1 at the 12-month
interview, while women with the fewest classes continued to show moderate decreases
in the number of unsafe events across time. Women who attended 6 to 15 classes
84

reported an increase in unsafe events at 12-months compared to 6-months. However,
because of the small number of individuals in this group it is difficult to draw
meaningful conclusions. In summary, women in both Treatment Groups demonstrated
significant declines in unsafe events over time. Seeking Safety attendance appeared to
moderate reductions at 6-months with greater reductions seen among women who
attended more classes and these reductions were maintained across the second six-
month time period.

Figure 3.9. Partial modeling results for unsafe events, Treatment Group, and Seeking
Safety attendance

Notes. * indicates significant parameter estimate, p < .05; t indicates trend, .10 < p <
.05; Group = treatment group; SSC
1
= number of Seeking Safety classes attended during
first three months of study; D
U3
= latent difference score for unsafe events between
baseline and 6-month interviews; D
U5
= latent difference score for posttraumatic stress
symptoms between 6-month interview and 12-month interviews.
85

Table 3.7. Means of Difference Scores for Unsafe Events by Treatment Group,
Interview Period, and Seeking Safety Classes.

Treatment Group
Total Classes

UE (6mo – BL)
Difference Scores

UE (12-mo – 6-mo)
Difference Scores

Treatment-as-Usual Group
0 classes
Mean = -2.0
SD = 2.9
n = 166

Mean = 0.0
SD = 2.0
n = 159

Integrated Treatment Group
0 - 5 classes
Mean = -1.0
SD = 3.2
n = 28

Mean = -1.0
SD = 2.3
n = 25

Integrated Treatment Group
6 – 15 classes
Mean = -1.3
SD = 1.3
n = 14

Mean = 0.6
SD = 1.6
n = 12

Integrated Treatment Group
16 or more classes
Mean = -2.8
SD = 2.1
n = 69

Mean = -0.1
SD = 1.2
n = 64

Notes. UE = number of different types of unsafe events reported in the previous six
months; 6-mo = 6-month interview; 12-mo = 12-month interview; N = number of
women providing valid data at each interview.

Additional descriptive analyses were conducted using the Use of Residential
Services variable described previously. Table 3.8 provides means of difference scores
for unsafe events by Treatment Group, Interview Period, and Residential Service Use
Group. Larger reductions in the frequency of different types of unsafe events during the
first six months of the study were seen among women in the High Residential Service
Use Groups (averaging nearly 3 events) compared to women in the Low Residential
Service Use Group (averaging 1 to 1.5 events) across levels of Treatment Group,
86

though there is a lot of variability within groups. There appeared to be little change in
unsafe events during the second half of the study among any of the groups. Thus, it
appears that the increased use of any residential treatment was related to reductions in
unsafe events during the first six months of the study; making it more difficult to clearly
attribute changes to participation in Seeking Safety.

Table 3.8. Means of Difference Scores for Unsafe Events by Treatment Group,
Interview Period, and Use of Residential Services Group.

Treatment Group
Total Classes

UE (6mo – BL)
Difference Scores

UE (12-mo – 6-mo)
Difference Scores

Treatment-as-Usual Group
Low Use of Residential
Services
Mean = -1.4
SD = 3.1
n = 71

Mean = -0.2
SD = 2.4
n = 70

Treatment-as-Usual
High Use of Residential
Services
Mean = -2.7
SD = 2.6
n = 85

Mean = 0.3
SD = 1.3
n = 80

Integrated Treatment Group
Low Use of Residential
Services
Mean = -0.9
SD = 2.3
n = 35

Mean = -0.6
SD = 1.9
n = 32

Integrated Treatment Group
High Use of Residential
Services
Mean = -2.9
SD = 2.2
n = 70

Mean = 0.0
SD = 1.4
n = 64

Notes. UE = number of different types of unsafe events reported in the previous six
months; 6-mo = 6-month interview; 12-mo = 12-month interview; N = number of
women providing valid data at each interview.

87

Days in residential treatment. Seven longitudinal growth curve models were
tested to model the trajectories for the number of days in residential treatment (see
Table 3.9). None of the basic models adequately described the data. Two
modifications were made to the non-linear Dual Change Model, which was the best-
fitting of the original models. In the Modified Dual Change Model (Model 5), two
changes in regards to residuals were made. First, the residuals for periods 1 and 2 were
allowed to vary while the residuals for period 3 – 5 were equated as they had been in all
previous models. This improved model fit and suggests the presence of different
subgroups. In addition, the outside factors influencing days in residential treatment
were different for the 90 days prior to women joining the study and during the first 3
months of the study compared to those in the subsequent 9 months of the study. The
second modification that was made to the residuals consisted of correlating residuals for
adjacent time periods for period 2 – 5 (i.e., correlating time periods 2 and 3, time
periods 3 and 4, and time periods 4 and 5). This suggests that factors influencing
women’s participation in residential treatment programs across the 12 months of the
study were temporally related. These statistical changes were retained in the final
model as they both improved model fit and conceptually could be understood (see
Figure 3.10 for depiction of final modeling results).
Inspection of the parameter estimates for the final model provides some insights
into the complexity of modeling the extent of women’s residential treatment across
time. The alpha parameters were equated across time and had a positive slope,

88
Table 3.9. Model Summary Table for Longitudinal Latent Difference Score Models of Days in Residential Treatment.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
1. No Change 1196.7 (17)
< .001

n/a n/a .00 15224 15236 .43 (.41, .45)
< .001
.95
2a. Proportional
Change
all betas equated

1189.1 (16)
< .001
1 7.6 (1)
< .05
.00 15219 15234 .45 (.42, .47)
< .001
.94
2b Proportional Change
all betas freed

497.9 (13)
< .001
1

2a

698.8 (4)
< .05

691.2 (3)
< .05
.00 14533 14561 .32 (.29, .34)
< .001
.54
3a. Constant Change
with alpha equated

461.4 (13)
< .001
1

735.5 (4)
< .05
.00 14497 14524 .31 (.28, .33)
< .001
.53

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df =
degrees of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison
model; CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion;
RMSEA = root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that
RMSEA is equal to or less than .05; SRMR = standardized root mean square residual.

89
Table 3.9. Continued.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
3b. Constant Change
with alpha equated

458.7 (11)
< .001
1

3a

738.0 (6)
< .05

2.7 (2)
> .05
.00 14498 14533 .33 (.31, .36)
< .001
.52
4. Dual Change

100.6 (9)
< .001

1

2b

3a

1096.1 (8)
< .05

397.3 (4)
< .05

360.8 (4)
< .05
.77 14144 14187 .17 (.14, .20)
< .001
.14

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df = degrees
of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison model; CFI
= Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; RMSEA = root
mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that RMSEA is equal to
or less than .05; SRMR = standardized root mean square residual.

90
Table 3.9. Continued.

Model

χ
2
(df)
p
Comparison
Model
Δ χ
2
(df)
p

CFI AIC BIC RMSEA (90% CI)
P(RMSEA < .05)
SRMR
5. Modified Dual Change
correlated residuals and
freed residuals

12.6 (6)
= .05
1

4

1184.1 (11)
< .05

98.0 (3)
< .05
.98 14062 14117 .06 (.00, .10)
.37
.03

Notes. * and bolded results signifies final model selected; χ
2
= chi square value associated with each model; df = degrees
of freedom; p = probability; Δ χ
2
= change in chi square values between current model and current comparison model;
CFI = Comparative Fit Index; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; RMSEA =
root mean square of approximation; 90% CI = 90% confidence interval; P(RMSEA < .05) = probability that RMSEA is
equal to or less than .05; SRMR = standardized root mean square residual.

91
Figure 3.10. Final modeling results for days in residential treatment

Notes. * indicates a significant parameter estimate, p < .05; L
Rx
= latent factors for days
in residential treatment at times 1 - 5; time 1 = baseline interview; time 2 = 3-month
interview; time 3 = 6-month interview; time 4 = 9-month interview; time 5 = 12-month
interview; D
Rx
= latent differences score at times 1 - 5; R
x
= raw scores at times 1 -5;
paths with no associated numeric were fixed to 1.0.

indicating that the number of days a woman spent in residential treatment tended to
increase across time especially for the first three month period; however, it cannot be
assumed this is time was spent in the original treatment program as the variable includes
the number of days receiving services in any residential treatment facility. The
implication is that once engaged in residential treatment women tended to stay in some
form of residential treatment, especially during the first three month time period.

92
However, this is offset by the proportional change part of the model, which showed that
during the year of the study, more days spent in treatment during each three-month
period predicted fewer days in treatment in the subsequent time study with the effect
growing stronger across time. Because the residential treatment variable summed
across residential sites, it cannot clearly be established if women were leaving due to
completing treatment, transferring between facilities or terminating care against medical
advice or otherwise. The mean for level indicates that women averaged about 23 days
in residential treatment in the three months prior to the study with little variation around
this point (SD = 11.1, p > .05). While there was a significant positive slope (about 9
days), there was a lot of variability around this estimate (84 days), but it was not
significant. Overall, there was a great deal of variability in days spent in residual
treatment, making these data hard to model even with complex nonlinear models.
However, as the model developed fit the data well, it was retained. In addition, as this
is serving as a control variable in this study, additional analyses identifying factors
influencing time spent in residential treatment were not conducted.
Next, the effects of Treatment Group and Seeking Safety attendance on changes
in days spent in residential treatment were added to the previous model (see Figure 3.11
for depiction of the final model). Inspection of parameter estimates indicated that both
Seeking Safety attendance variables and Treatment Group had significant predictive
effects on subsets of the latent difference scores. Greater frequency of Seeking Safety
attendance during the first three months of the study predicted more time spent in

93
Figure 3.11. Partial modeling results for days in residential treatment with
Treatment Group and Seeking Safety attendance

Notes. * indicates a significant parameter estimate; Group = treatment group; SSC
1
=
number of Seeking Safety classes taken by women in the Integrated treatment group
during first three months of the study; SSC
2
= number of Seeking Safety classes taken
by women in the Integrated treatment group during the second three months of the
study; D
Rx
= latent difference score for days in residential treatment at times 1 -5; time 1
= baseline interview; time 2 = 3-month interview; time 3 = 6-month interview; time 4 =
9-month interview; time 5 = 12-month interview.

residential treatment agencies at the first three time points (i.e., 0 – 3 months; 4 – 6
months, and 7 – 9 months), but not during the last three months of the study (10 – 12
months). This effect is not necessarily remarkable at the first time period, because
women in the Integrated Treatment Group had to be in their original residential
treatment facility to attend Seeking Safety classes. However, Seeking Safety attendance
in the first three months of the study predicting more days spent in residential treatment
during months 4 – 6 and months 7 -9 suggests that greater engagement in the Seeking

94
Safety program early on predicted more time receiving residential services overall, even
after Seeking Safety classes ended. Seeking Safety attendance during the second three
months of the study also predicted more use of residential treatment services at three
time points (months 3 -6, months 7 – 9, and months 10 -12).
Treatment Group predicted latent difference scores at the first two three-month
periods. Similar to unsafe events, there appears to be an interaction-like effect between
Treatment Group and Seeking Safety attendance. Table 3.10 displays the raw means
for days spent in residential treatment, broken down by Treatment Group, Interview
Period and Seeking Safety attendance. Among women in the Integrated Treatment
Group, those who attended the fewest Seeking Safety classes also left treatment earlier
compared to women with more Seeking Safety classes, which is consistent with women
having to be in their original residential treatment facility to attend Seeking Safety.
When the model statistically controlled for the effect of Seeking Safety attendance on
days in residential treatment, this resulted in the effect seen that women in the
Treatment-as-Usual Group spent significantly more days in residential treatment
compared to women in the Integrated Treatment Group at the first two time periods.
Inspection of overall model fit indicated that this model fit the data somewhat poorly.
RMSEA (.09) and SRMR (.06) were both below expected values, while CFI indicated
good model fit.
Again, toward teasing apart the effects of Seeking Safety attendance from time
spent in residential treatment on the frequency of subsequent levels of time spent in

95
residential treatment, additional analyses using the Residential Service Use Group
variable were conducted. Table 3.11 displays means for days in residential by
Treatment Group, Interview Period, and Residential Service Use Group. Not
unexpectedly, there is little difference among the groups for use of residential treatment
services in the three months prior to the start of the study. Large differences in time
spent in residential treatment based on Residential Service Use Group for times periods
covering 0 to 3 months and 4 to 6 months also are not remarkable as the Residential
Service Use Groups were created by grouping women according to who attended at
least 130 days of residential treatment during the first six months of the study or not.
The final two columns of Table 3.11 are interesting. Regardless of Treatment Group,
women in the Low Residential Service Use Groups, on average, did not participate in
any type of residential treatment during the second half of the study. However, among
those in the High Residential Service Use Groups, women from the Integrated
Treatment Group tended to report spending more days in residential treatment during
months 6 – 9 and months 10 – 12 compared to women in the Treatment-as-Usual Group
for the same time periods. Based on both sets of analyses, it may be that Seeking Safety
attendance helped women to remain engaged in some type of residential treatment
longer than those who were not exposed to Seeking Safety, though it cannot be clearly
determined that the women remained in their original treatment program and there
remains significant variability in the number of days women spent in residential
treatment.

96
Table 3.10. Raw Means for Days in Residential Treatment Settings by Treatment Group, Interview Period, and
Seeking Safety Attendance.

Treatment Group
Total Classes

3 months prior
to Baseline
Baseline to
3-months

3-months to
6-months

6-months to
9-months

9-months to
12-months
Treatment-as-Usual
0 classes
Mean = 26.3
SD = 19.6
n = 215

Mean = 74.9
SD = 25.2
n = 186
Mean = 48.5
SD = 37.3
n = 170
Mean = 24.0
SD = 34.4
n = 161
Mean = 13.1
SD = 26.9
n = 169
Integrated Group
0 - 5 classes
Mean = 17.9
SD = 10.1
n = 43

Mean = 41.0
SD = 28.9
n = 29
Mean = 22.6
SD = 36.0
n = 25
Mean = 13.5
SD = 31.0
n = 26
Mean = 14.5
SD = 30.6
n = 28
Integrated Group
6 – 15 classes
Mean = 17.0
SD = 17.6
n = 33

Mean = 80.4
SD = 13.1
n = 32
Mean = 42.3
SD = 39.9
n = 15
Mean = 17.6
SD = 32.5
n = 13
Mean = 9.7
SD = 27.4
n = 11
Integrated Group
16 or more classes
Mean = 18.4
SD = 15.4
n = 79

Mean = 87.8
SD = 9.7
n = 79
Mean = 80.8
SD = 21.9
n = 71
Mean = 71.7
SD = 34.2
n = 63
Mean = 43.6
SD = 41.3
n = 64

97
Table 3.11. Raw Means for Days in Residential Treatment Settings by Treatment Group, Interview Period, and Use
of Residential Services Group.

Treatment Group

3 months prior
to Baseline
Baseline to
3-months

3-months to
6-months

6-months to
9-months

9-months to
12-months
Treatment-as-Usual Group
Low Use of Residential
Services
Mean = 23.1
SD = 15.8
n = 73

Mean = 62.2
SD = 29.4
n = 73
Mean = 12.6
SD = 18.7
n = 73
Mean = 7.0
SD = 20.4
n = 67
Mean = 4.1
SD = 17.5
n = 65
Treatment-as-Usual Group
High Use of Residential
Services
Mean = 27.0
SD = 20.5
n = 89

Mean = 88.2
SD = 8.7
n = 89
Mean = 80.2
SD = 14.7
n = 89
Mean = 37.6
SD = 37.6
n = 79
Mean = 17.1
SD = 28.6
n = 82
Integrated Treatment Group
High Use of Residential
Services
Mean = 16.5
SD = 11.1
n = 36

Mean = 55.4
SD = 29.8
n = 36
Mean = 12.9
SD = 22.7
n = 36
Mean = 11.1
SD = 29.0
n = 30
Mean = 12.8
SD = 30.1
n = 30
Integrated Treatment Group
High Use of Residential
Services
Mean = 20.0
SD = 16.0
n = 74

Mean = 88.8
SD = 7.2
n = 74
Mean = 86.7
SD = 9.5
n = 74
Mean = 72.8
SD = 32.3
n = 63
Mean = 44.5
SD = 40.8
n = 61

98
Hypothesis Testing
Two sets of hypotheses were tested. The first set of hypotheses assessed
whether the Seeking Safety attendance predicted lower posttraumatic stress symptoms
either during either 1a) the first six-month period of the study when many women were
in the original treatment settings or 1b) the second six-month period of the study when
most women were no longer in the original treatment settings. Specifically, it was
hypothesized that greater participation would predict reductions in posttraumatic stress
symptoms after controlling for WCDVS Treatment Group, the potential increases in
posttraumatic stress symptoms associated with the occurrence of unsafe events and the
number of days each woman spent in residential treatment during each three-month
period. The second pair of hypotheses was similar except it assessed the impact of
Seeking Safety attendance on unsafe events. Specific hypotheses assessed whether the
number of Seeking Safety classes a woman attended predicted significant reductions in
the number of different types of unsafe events a woman reported 2a) during the first six-
month period of the study or 2b) during the second six-month period, after controlling
for Treatment Group, posttraumatic stress symptoms, and residential treatment.
Hypotheses 1a and 1b. Overall, the final model used for hypotheses testing
reproduced the data well; χ2 (43) = 46.2, p = .34; AIC = 32483; BIC = 32910; CFI =
1.00; RMSEA = .01 (00, .04); probability (RMSEA < .05) = 1.00; SRMR = .02. Tables
3.12a – 3.12f summarize the modeling results for the final model, which included the
longitudinal models of posttraumatic stress symptoms, unsafe events, and days in

99
residential treatment, with all cross-lagged effects as well as the three external variables,
Treatment Group, Seeking Safety attendance during the first three-month time period
and Seeking Safety attendance during the second three-month time period. Inspection
of the column for posttraumatic stress symptoms parameter estimates in Tables 3.12d
and 3.12e (columns labeled “PSS”) shows that both hypotheses 1a and 1b were not
supported. When statistically controlling for the potential impact of additional unsafe
events, Treatment Group, and days spent in secured, residential treatment settings,
neither of the Seeking Safety attendance variables predicted significant changes in
posttraumatic stress symptoms at any time point
6
.
There are a number of factors that could mask any potential effects of Seeking
Safety attendance on posttraumatic stress symptoms. First, as discussed throughout the
results section, there is a large of heterogeneity in the data across time making modeling
the data difficult, especially for days in residential treatment. Second, there was
difficulty finding an appropriate control group for testing effects of Seeking Safety in
this dataset as Seeking Safety attendance was confounded with days spent in residential
treatment services. The total number of Seeking Safety classes attended correlated
significantly with days in residential treatment at months 0 – 3 ( ρ = .6, p < .05), months
4 -6 ( ρ = .7, p < .05), months 7 – 9 ( ρ = .6, p < .05) and months 10 -12 ( ρ = .4, p < .05).
In Table 3.4, means for women in the Integrated Treatment Group showed small

6
Final hypotheses testing analyses were repeated on the complete PSS-SR scale (not shown) and results
were not substantially different in relation to study hypotheses or underlying latent growth curve model.
Results using current modified version of PSS-SR were used as they had the benefit of longitudinal
invariance.

100
differences in mean scores compared to the variability with group for women who
attended different levels of Seeking Safety (i.e., 0 -5 sessions, 6 – 15 sessions, and 16 or
more sessions). For this comparison, it is unclear what additional treatment services
women may have received to account for the reduction in posttraumatic stress
symptoms seen in the groups; though it is known that women in the Treatment-as-Usual
Group could have participated in a small number of domestic violence classes offered in
that Treatment Group. Descriptive results comparing women who used more residential
treatment services during the first six months of the study compared to those who spent
fewer days in residential treatment during the same period (Table 3.5) showed a large
drop in posttraumatic stress symptoms among women in the Treatment-as-Usual Group
who participated in higher levels of residential treatment compared to the other groups,
because there was still significant variability within the group. No other differences in
scores were apparent. Thus, while posttraumatic stress symptoms scores did tend to
decrease across time, it remains unclear what is influencing these symptom reductions.
Hypotheses 2a and 2b. Tables 3.12d and 3.12e provide the parameter estimates
necessary to assess this pair of hypotheses. In the column for unsafe events (labeled
“UE”), there is one significant parameter estimate in Table 3.12d indicating that
Seeking Safety attendance during the first three months of the study predicted lower
levels of unsafe events during the first six months of the study (parameter estimate = -
0.1, p < .05), even after averaging out effects for Treatment Group, posttraumatic stress
symptoms, and days in residential treatment. However, Seeking Safety attendance

101
during the second three months of the study did not predict systematic changes in
unsafe events (parameter estimate = -0.1, p > .05; Table 3.5e). Thus, Hypothesis 2a
was partially supported; in that women who attended more Seeking Safety classes
during the first three months of the study reported significant decreases in unsafe events
during the first six months of the study. However, this effect must be interpreted
cautiously, as we know based on earlier results that women in the Treatment-as-Usual
group showed similar reductions in unsafe events during the first six months of the
study. In addition, descriptive statistics based on level of residential service use
suggested similar reductions in unsafe events for women in the high residential service
use category independent of Treatment Group compared to women in the low
residential service use category. Thus, reductions in unsafe events may be related to a
woman’s presence in a secured environment or that safe coping skills were acquired in
other residential facilities or a combination of these and other factors.
Hypothesis 2b was not supported with neither of the Seeking Safety attendance
variables predicting significant changes in unsafe events during the second six months
of the study; though it is important to remember that women appeared to retain
treatment gains made during the first six months of the study. In addition, neither of the
Seeking Safety attendance variables was correlated with initial status, such that the
number of different types of unsafe events a woman experienced during the six-month
period prior to the beginning of the study did not predict how many Seeking Safety
classes she attended during the first six months of the study.

102
Power associated with final model. Determining power related to latent growth
curve model, especially for power to detect non-zero parameter estimates, is a largely
unstudied area (Duncan, Duncan, & Strycker, 2006; Preacher, Wichman, MacCallum,
& Briggs, 2008). A power estimate on the final model was conducted based on the
methodology of overall model fit suggested by MacCallum, Browne and Sugawara
(1996). In this process, null and alternative hypotheses are chosen representing values
of model fit as represented by RMSEA that reflect good fit ( ε
0
) and poor fit ( ε
A
).
Values of ε
0
= .05 and ε
A
= .08 were chosen for the current analysis, representing “close
fit” and “mediocre fit”, respectively (Preacher et al., 2008). In this methodology, the
null hypothesis is reversed in regards to standard hypothesis testing procedures and is
associated with the not-good fit model. Thus, if rejected, significant results provide
strong support for good model fit (MacCallum et al., 1996). Using alpha = .05, degrees
of freedom = 43, sample size = 370, power was estimated to be .95 in the final model.
Methods to assess power for detecting group differences in linear models (Fan, 2003),
for slope coefficients in single indicator models (Hertzog, von Oertzen, Ghisletta, &
Lindenberger, 2008), and estimation of power for treatment effects in experimental
studies (Duncan, Duncan, Strycker, & Li, 2002) were identified, but could not be used
as the current model included non-linear growth parameters, study design was quasi-
experimental with non-randomization to treatment groups and hypothesis testing
focused on cross-lagged association instead of slope parameters.

103
Table 3.12a: Modeling Results for Final Model: Latent Growth Curve Model
Parameter Estimates

M / Var
/ Corr

PSS UE RES

Initial status mean

n/a 15.5* 3.7* 22.8*
Initial status variance

n/a 30.8* 4.1* 49.5*
Constant Change Mean

n/a -0.2 -1.6 0.7
Constant Change Variance

n/a 0.6 4.6* 14.4
Constant Change Coefficients
Alpha
1
to LDS
time 1
(all fixed to 1)
Alpha
2
to LDS
time

2

Alpha
3
to LDS
time

3

Alpha
4
to LDS
time

4

n/a
1.00
f

2.9
1

2.9
1

-2.3

n/a
1.00
f

n/a
-.2
t

1.00
f

4.6
2

4.6
2

4.6
2

Initial Status with Constant Change

n/a 0.1 -2.9 3.3
Proportional Change Effects (Days in Residential Treatment only)

RES
1
→ LDS
RES2

RES
2
→ LDS
RES3

RES
3
→ LDS
RES4

RES
4
→ LDS
RES5

n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
3.4*
-.6*
-.8*
-1.0*
Residual Measurement Error (all times)
time 1
time 2
time 3 -5
n/a 33.4* 1.5* n/a
279.4*
293.8*
713.7*
3

Notes. M = mean; Var = variance; Corr = correlation; PSS = posttraumatic stress
symptoms; UE = unsafe events; * and bolded = critical ratio test associated with
parameter estimate significant p < .05; t = critical ratio test associated with parameter
estimate .10 < p < .05 (trend); f = parameter fixed to specified value; parameters with
same numerical superscript were equated; RES
1-5
= latent scores for days in residential
treatment; time 1 – time 5 = interviews at baseline, 3-months, 6-months, 9-months, and
12-months.

104
Table 3.12b: Modeling Results for Final Model: Cross-lagged Associations Parameter
Estimates

Cross-lagged Associations

M / Var
/ Corr

PSS UE RES

PSS
1
→ LDS
UE1,
LDS
RES1
n/a n/a -.1 -.2

PSS
2
→ LDS
UE2
, LDS
RES2

n/a n/a -.4* 1.2
PSS
3
→ LDS
UE2
, LDS
RES3

n/a n/a .4* 1.6
PSS
4
→ LDS
RES4

n/a n/a n/a 0.8
UE
1
→ LDS
PSS1
, LDS
RES1

n/a .1
3
n/a -10.4*
UE
2
→ LDS
PSS2
, LDS
RES2
,
LDS
PSS3,
LDS
RES3

n/a
n/a

.1
3

.1
3

n/a
n/a
-.6
-2.3
UE
3
→ LDS
PSS4
, LDS
RES4

n/a .1
3
n/a .5
RES
1
→ LDS
UE1

n/a n/a 0.0
4
n/a
RES
2
→ LDS
PSS1
, LDS
UE1

n/a -0.1* 0.0
4
n/a
RES3 → LDS
PSS2
, LDS
UE2

n/a 0.0 0.0
4
n/a
RES4 → LDS
PSS3
, LDS
UE3

n/a 0.0 0.0
4
n/a
RES5 → LDS
PSS4

n/a 0.1 n/a n/a

Notes. M = mean; Var = variance; Corr = correlation; PSS = posttraumatic stress
symptoms; UE = unsafe events; * and bolded = critical ratio test associated with
parameter estimate significant p < .05; t = critical ratio test associated with parameter
estimate .10 < p < .05 (trend); f = parameter fixed to specified value; parameters with
same numerical superscript were equated; RES = days in residential treatment; RES
1-5
=
latent scores for days in residential treatment; LDS = latent difference score; time 1 –
time 5 = interviews at baseline, 3-months, 6-months, 9-months, and 12-months; LDS
time 1
– LDS
time 5
= column title specifies latent difference score at times 1 - 5.

105
Table 3.12c: Modeling Results for Final Model: Parameter Estimates for Effects of
Treatment Group on Dependent Variables

Treatment Group Effects

M / Var /
Corr

PSS UE RES

Treatment Group → LDS
time 1

n/a 0.6 n/a 11.6
Treatment Group → LDS
time 2

n/a -0.3 1.3 -21.4*
Treatment Group → LDS
time 3

n/a -2.6 n/a -11.9
Treatment Group → LDS
time 4

n/a -1.6 -0.1 -0.5
Treatment Mean

0.4* n/a n/a n/a
Treatment Variance

0.2* n/a n/a n/a
Treatment Group with Initial Status

n/a 0.1 0.0 -2.0*
Treatment Group with Seeking Safety 1

2.5* n/a n/a n/a
Treatment Group with Seeking Safety 2

0.3* n/a n/a n/a

Notes. M = mean; Var = variance; Corr = correlation; PSS = posttraumatic stress
symptoms; UE = unsafe events; * and bolded = critical ratio test associated with
parameter estimate significant p < .05; t = critical ratio test associated with parameter
estimate .10 < p < .05 (trend); f = parameter fixed to specified value; parameters with
same numerical superscript were equated; RES = days in residential treatment; RES
1-5
=
latent scores for days in residential treatment; LDS = latent difference score; time 1 –
time 5 = interviews at baseline, 3-months, 6-months, 9-months, and 12-months; LDS
time 1
– LDS
time 5
= column title specifies latent difference score at times 1 - 5.

106
Table 3.12d: Modeling Results for Final Model: Parameter Estimates for Effects of
Number of Seeking Safety Classes Taken during Months One through Three on
Dependent Variables

First Three Months of Seeking Safety
(Seeking Safety 1) Effects

M / Var
/ Corr
PSS UE RES
Seeking Safety 1 → LDS
time 1

n/a 0.1 n/a 2.3*
Seeking Safety 1 → LDS
time 2

n/a 0.1 -0.1* 0.9*
Seeking Safety 1 → LDS
time 3

n/a 0.1 n/a 1.3*
Seeking Safety 1 → LDS
time 4

n/a -0.1 0.0 0.7
Seeking Safety 1 Mean

4.2* n/a n/a n/a
Seeking Safety 1 Variance

50.1* n/a n/a n/a
Seeking Safety 1 with Initial Status

n/a -1.4 0.2 -21.9*
Seeking Safety 1 with Seeking Safety 2

.5* n/a n/a n/a
Notes. M = mean; Var = variance; Corr = correlation; PSS = posttraumatic stress
symptoms; UE = unsafe events; * and bolded = critical ratio test associated with
parameter estimate significant p < .05; t = critical ratio test associated with parameter
estimate .10 < p < .05 (trend); f = parameter fixed to specified value; parameters with
same numerical superscript were equated; RES = days in residential treatment; RES
1-5
=
latent scores for days in residential treatment; LDS = latent difference score; time 1 –
time 5 = interviews at baseline, 3-months, 6-months, 9-months, and 12-months; LDS
time 1
– LDS
time 5
= column title specifies latent difference score at times 1 - 5.

107
Table 3.12e: Modeling Results for Final Model: Parameter Estimates for Effects of
Number of Seeking Safety Classes Taken during Months Four to Six on Dependent
Variables

Second Three Months of Seeking
Safety (Seeking Safety 2) Effects

M / Var
/ Corr
PSS UE RES
Seeking Safety 2 → LDS
time 2

n/a -0.1 -0.1 2.7*
Seeking Safety 2 → LDS
time 3

n/a 0.1 n/a 2.2*
Seeking Safety 2 → LDS
time 4

n/a -0.1 0.0 1.6*
Seeking Safety 2 with Initial Status

n/a -2.7
t
-0.2 1.0
Seeking Safety 2 Mean

0.6* n/a n/a n/a
Seeking Safety 2 Variance

18.2* n/a n/a n/a
Notes. M = mean; Var = variance; Corr = correlation; PSS = posttraumatic stress
symptoms; UE = unsafe events; * and bolded = critical ratio test associated with
parameter estimate significant p < .05; t = critical ratio test associated with parameter
estimate .10 < p < .05 (trend); f = parameter fixed to specified value; parameters with
same numerical superscript were equated; RES = days in residential treatment; RES
1-5
=
latent scores for days in residential treatment; LDS = latent difference score; time 1 –
time 5 = interviews at baseline, 3-months, 6-months, 9-months, and 12-months; LDS
time 1
– LDS
time 5
= column title specifies latent difference score at times 1 - 5.

108
Table 3.12f. Modeling Results for Final Model: Correlated Growth Factors.

Level PSS

Slope PSS Level UE Slope UE Level RES
Slope PSS

0.1
Level UE

4.3* -0.5
Slope UE

1.3 0.4 -2.9
Level RES

5.7 -0.2 9.8* -9.5
t

Slope RES

-2.5 -1.3 0.9 -1.8 3.3
Notes. PSS = posttraumatic stress symptoms; UE = unsafe events; RES = days in
residential treatment; * and bolded signifies significant covariance, p < .05;
t
signifies
trend, .10 < p < .05.

Interpretation of Final Modeling Results
Predictors of posttraumatic stress symptoms. Inspection of columns marked
“PSS” in Tables 3.12a – 3.12e describe the final modeling effects for posttraumatic
stress symptoms in the context for unsafe events, days in residential treatment,
Treatment Group and both Seeking Safety attendance variables. Table 3.12a
summarizes the results of the latent growth curve model of posttraumatic stress
symptoms. Results indicate that initial levels of posttraumatic stress symptoms were
significantly different from zero (Mean = 15.5) and that there was significant variability
around this mean value (30.8). Interestingly, the slope associated with this Constant
Change Model was no longer significant when all other variables were added to the

109
model. In addition, the alpha parameters, depicting natural changes in posttraumatic
stress symptoms across time, were also no longer significant in the final model. Initial
univariate modeling results indicated that proportional change parameters did not
significantly add to the model and were not included
7
. Thus, women did report
important levels of posttraumatic stress symptoms and there was significant variability
in these symptom levels across time. However, neither of the internal change
parameters proposed (i.e., those assessing either natural or proportional change)
significantly predicted symptom change scores when all other variables were added to
the model. Table 3.12b documents the cross-lagged effects in the final model. Only
one parameter significantly predicted changes in posttraumatic stress symptoms.
Specifically, the more days spent in residential treatment settings during the first three
months of the study predicted significant decreases in posttraumatic stress symptoms at
the three-month interview. This is interesting as it was expected that posttraumatic
stress symptoms might increase during the initial months of treatment when reduced
avoidance of symptoms through substance use and other negative coping skills as well
as increases in symptoms related to physiological withdrawal symptoms may have
temporarily increased posttraumatic stress symptom levels. Table 3.12c summarizes the
effects of Treatment Group on posttraumatic stress symptoms, and shows that
Treatment Group did not predict significant changes in posttraumatic stress symptoms
at any time point when all other variables were also in the model. Thus, there was a

7
Additional analyses (not reported here) added the proportional change parameters to the final model and
resulted in no additional model fit or significant predictive effects with their inclusion.

110
significant reduction in posttraumatic stress symptoms during the first three months of
the study that appeared to be maintained across time. Interestingly, time spent in
general residential treatment predicted the significant decreases in posttraumatic stress
symptoms instead of the hypothesized Seeking Safety attendance.
Predictors of unsafe events. The columns titled “UE” in Tables 3.12a – 3.12e
summarize the modeling results for unsafe events in the final model. The mean level of
unsafe events was significantly different from zero, with women reporting an average
3.7 events in the six months prior to the beginning of the study. In addition, there was
significant variability in the number of different types of unsafe events women reported
(SD of initial level = 4.1). In the Constant Change Model incorporated into the
complex final model, the slope coefficient (representing natural change in scores) was
no longer significant in the context of the other variables. However, the variability
around the slope was significant, indicating that women differed significantly in the rate
of changes in unsafe events across time. In addition, there was a trend toward a
significant increase in unsafe events during the second six-month time period of the
study (i.e., negative constant change mean multiplied by negative alpha coefficient).
The mean and median number of unsafe events reported at 12-months for both
Treatment Groups were approximately zero, as there had been a large reductions in
unsafe events reported at 6-months, representing a near floor effect. Thus, women
tended to retain any gains made during the first six months of the study.

111
Investigation of the cross-lagged effects of posttraumatic stress symptoms and
days in residential treatment on unsafe events (Table 3.12b) showed that posttraumatic
stress symptoms during the second and third three-month time periods predicted
significant changes in the number of different types of unsafe events reported during the
second six months of the study. However, the effects were in the opposite directions.
Higher levels of posttraumatic stress symptoms during months three to six of the study
predicted significant decreases in unsafe events during the last half of the study, while
higher levels of posttraumatic stress symptoms during months six to nine of the study
predicted significant increases in unsafe events.
Predicting days in residential treatment. While included as a control variable,
results of the impact of other study variables on days spent in residential treatment are
still included in Tables 3.12a – 3.12e and can provide some insight into use of
residential treatment services across time among women with co-occurring disorders.
Table 3.12a describes the Dual Change Model included in the final model. The initial
mean was significantly different from zero (mean = 22.8), indicating that women spent
about three weeks in residential treatment settings during the three-month period prior
to the start of the study. The significant variance around the mean level suggests there
was significant variability in the amount of time women reported spending in residential
treatment programs prior to beginning the current study. As with the previous two
models, the mean slope did not differ significantly from zero, indicating no systematic
natural change across time in this model. In addition, the alpha coefficients were not

112
significant once the effects of other study variables were included in the model.
However, proportional change coefficients (the effects of where a woman started on the
variable at the previous time point on the current time point) were significant at all
occasions. The first beta coefficient was positive reflecting the large increase in the
number of days women spent in residential treatment during the first three months of
the study. The next three coefficients grow increasingly larger with a negative sign,
indicating that the more days women spent in treatment during each time period
predicted significantly fewer days in the following three-month time periods. Figure
3.5 may provide insight into the nature of these effects. Inspection of the data suggest
that women who reported more days in residential treatment during the first three
months of the study tended to remain higher for a time period or two, then their
residential treatment use dropped precipitously for the remaining time periods; perhaps
as then completed residential treatment or discontinued treatment and did not start at
any other agency. Identification of subpopulations based on program completion may
be useful in future studies. Table 3.12b summarizes the cross-lagged effects of
posttraumatic stress symptoms and unsafe events on days in residential treatment.
There was only one significant cross-lagged effect, with women reported more unsafe
events in the six-months prior to the beginning of the study spending significantly fewer
days spent in residential treatment during the first three months of the study compared
to women who reported fewer unsafe events during the six months prior to the
beginning of the study. This suggests that women who reported more different types of

113
stressful and traumatic events in the months just prior to the treatment study tended to
spend fewer days in residential treatment settings during the first three months of the
study. Inspection of the effects of Treatment Group on days in residential treatment
showed that when the model statistically accounted for all other study variables,
Treatment Group only predicted significant changes in days in residential treatment
during the second three months of the study with women in the Integrated Treatment
Group reporting significantly fewer days in residential treatment compared to women in
the Treatment-as-Usual Group. This may be counterintuitive since raw data suggest
that women in the Integrated Treatment Group reported more days in residential
treatment compared to women in Treatment-as-Usual Group. However, the effect is
likely a statistical artifact, similar to the effects seen in the univariate analyses for
unsafe events. In the Integrated Treatment Group, greater Seeking Safety attendance
predicted more days in residential treatment while any increases in time spent in
residential treatment for women in the Treatment-as-Usual Group were ascribed to the
Treatment Group variable. The significant correlation between Treatment Group and
initial status reflects for previously reported results that women in the Treatment-as-
Usual Group reported significantly more days in residential treatment settings in the
three months prior to the beginning of the study compared to women in the Integrated
Treatment Group. As seen in Tables 3.12d and 3.12e, the Seeking Safety attendance
variables are the most related to changes in days in residential treatment, as would be
expected since women would have had to remain in residential treatment to attend

114
Seeking Safety classes. As expected, more days in residential treatment during the first
three months of the study were associated with attending more Seeking Safety classes
during the first three months of the study. In addition, greater Seeking Safety
attendance during the first three months of the study predicted significantly more days
in residential treatment during months three to six and months six to nine compared to
those who attended fewer Seeking Safety classes. A similar pattern was seen with
Seeking Safety attendance during the second three months of the study with greater
attendance predicting more days in residential treatment during the last three time
periods.
There was a significant negative correlation between initial level of days in
residential treatment (i.e., number of days spent in residential treatment in the three
months prior to the study) and Seeking Safety attendance during the first three months
of the study. It is unclear why women who had been more residential treatment prior to
joining the WCDVS would take fewer Seeking Safety classes during the first three
months. However, there also was a trend for a negative correlation between initial
levels of residential treatment and slope for residential treatment. This indicates that on
average more days in residential treatment during the three months prior to the start of
the study predicted larger, negative slopes – or these women tended to leave treatment
earlier than those with fewer days. This may represent some women hopping from
treatment program to treatment program without completing them.

115
Table 3.12f displays the correlated growth factors for the three longitudinal
variables in the model. Two significant correlations were found. As one would expect,
more unsafe events reported in the six months prior to the start of the study was related
to significantly higher levels of posttraumatic stress symptoms at baseline. Also, more
days in residential treatment in the three months prior to the start of the study was
related to significantly higher reports of unsafe events in the six months prior to the start
of the study. However, there is no way to determine the timing or relationship between
these unsafe events and residential treatment usage; whether women who experienced
more unsafe events perhaps did not engage in residential treatment, whether unsafe
events were happening within residential treatment facilities leading women to leave
treatment, or whether women who experienced more unsafe events perhaps started and
stopped treatment more quickly than women who experienced fewer unsafe events.

116
Chapter 3: Discussion
Using the Los Angeles site data of the WCDVS, this study investigated the
impact of whether participation in and the level of participation in a trauma-focused
treatment program predicted changes in two trauma-related outcomes among women
with co-occurring disorders and histories of interpersonal violence who were already
participating in a treatment outcome study assessing the effects of a larger package of
integrated mental health and substance abuse services. Across the 12 months of the
study, women showed significant decreases in posttraumatic stress symptoms and
unsafe events. However, contrary to the first pair of hypotheses – those focusing on
posttraumatic stress symptoms as the main outcome – when statistically accounting for
the impact of WCDVS treatment group, time spent in residential treatment, previous
levels of posttraumatic stress symptoms and unsafe events, attendance in the trauma-
focused treatment program, Seeking Safety (Najavits, 2002), did not predict reductions
in posttraumatic stress symptoms at any of time point. However, the impact of trauma-
focused treatment on changes in posttraumatic stress symptoms was difficult to assess
clearly due to high levels of heterogeneity in the dataset for several of the longitudinal
variables, difficulty establishing a clear comparison group, and confounding of trauma-
focused treatment with time spent in residential treatment. However, even when
restricting analyses to investigating changes in posttraumatic stress symptoms based on
amount of trauma-focused treatment within only the WCDVS treatment arm that
received the integrated services package (including trauma-focused treatment), the

117
amount of trauma-focused treatment did not predict significant reductions in
posttraumatic stress symptoms. Though, heterogeneity in the data made model
development difficult, and future studies could investigate exogenous variables
predicting different courses of posttraumatic stress symptoms across time.
The second set of hypotheses (those focusing on unsafe events as the main
outcome) was partially supported, in that greater attendance in the trauma-focused
therapy predicted significant decreases in unsafe events among women receiving
integrated mental health and substance abuse services during the first six months of the
study. However, women in the comparison group, who did not receive the trauma-
focused treatment classes, also showed similar significant reductions in unsafe events
during the first six months of the study. Descriptive analyses focusing on the amount of
time women spent in any residential treatment setting suggested that reductions in the
frequency of unsafe events may be related to residential treatment in general rather than
trauma-focused treatment specifically. Contrary to the last hypothesis, level of
attendance in trauma-focused treatment did not predict further reductions in unsafe
events in the second half of the study; however, the treatment gains from the first half of
the study were maintained during the second six-months of the study. There, indeed,
appeared to be a floor effect for many of the women in regards to number of unsafe
events. After averaging nearly four different types of stressful and/or traumatic events
at baseline, the median number of such events at the 6-month and 12-month interviews

118
was zero. Also, women in the comparison group also tended to maintain their safety
gains.
Posttraumatic stress symptoms. Based on the final univariate model of
posttraumatic stress symptoms, significant reductions in symptoms were seen at 3-
months, 6-months, and 9-months for all women with larger decreases seen in the first
three-months of the study, at which time a majority of women remained in their original
treatment program. Previous studies using data from the WCDVS found significant
reductions in posttraumatic stress symptoms at 12-months in integrated treatment group
compared to the comparison group (Gatz et al., 2007; Gatz et al., 2004; Morrissey,
Jackson et al., 2005) and, more recently, showed that women with the most severe
levels of posttraumatic stress symptoms and substance abuse and who received
integrated services showed the greatest reductions in posttraumatic stress symptoms
(Cusack, Morrissey, & Ellis, 2008; Ellis & Morrissey, 2009). Analyses with only
posttraumatic stress symptoms and WCDVS treatment group in the current study
replicated those results (results not shown). Failure to replicate the findings in
multivariate models may be due to the presence of several correlated constructs in the
same model vying for the same variance pool with each effect size being lessened as
well as the factors previously discussed.
Only one variable significantly predicted reductions in posttraumatic stress
symptoms across time and only at one time point when all study variables were
included in the model. Specifically, more days in residential treatment during the first

119
three-months of the study predicted significant decreases in posttraumatic stress
symptoms at the three-month interview, while trauma-focused treatment did not. This
suggests that engagement in the broader treatment programs, and not the trauma-
focused group specifically, during acute treatment may predict reductions in
posttraumatic stress symptoms. However, because the trauma-focused treatment was an
integral part of the integrated treatment group, the potential benefits of trauma-focused
treatment cannot be ruled out based on these data. Women in the integrated services
group participated in an average of 10 of the 31 available sessions of Seeking Safety
classes during the first three-months of the study. This number of sessions may not
have provided sufficient opportunity for acquiring and/or the practice of skills that
strongly affect posttraumatic stress symptom. Yet, previous studies of the Seeking
Safety protocol have shown significant reductions in symptom levels with a minimal six
sessions required for completion (Hien et al., 2004; Najavits et al., 1998), though,
studies also show that often participants remain engaged in the treatment and take most
of the sessions offered (Gatz et al., 2004; Najavits et al., 2005; Najavits et al., 1998).
Analyses of total time spent in residential treatment suggested that women in the
integrated treatment group were in residential treatment for more days across the year
study period compared to comparison group, especially during later time periods. This
could be a function of differences in lengths of treatment programs favoring longer
treatment programs in the integrated treatment group or that women who participated in
the trauma-focused treatment tended to remain engaged in their residential program

120
longer than those who did not participate in residential treatment. Finally, the absence
of relationship between posttraumatic stress symptoms and trauma-focused treatment
could be caused by the dissemination of trauma-informed services both within the
integrated treatment site and within treatment programs, in general, which may have
provided women with some basic coping skills. In addition, their presence in controlled
facilities may have provided some increased sense of safety that may have affected
posttraumatic stress symptoms during the earliest phase of treatment.
This study found little evidence for a significant impact of a woman’s level of
posttraumatic stress symptomatology on time spent in residential services or
participation in trauma-focused treatment. Only when modeling data from the
integrated treatment group alone did higher levels of baseline posttraumatic stress
symptoms predict significant reductions in the amount of trauma-focused treatment and
only in the months four through six of the study. Thus, concerns that individuals may
disengage from treatment due to elevated psychological distress was not generally
supported in this study. Other studies of dually-diagnosed individuals receiving
integrated treatment found that those with higher posttraumatic stress symptoms,
especially avoidance symptoms were less likely to complete treatment (Brady, Dansky,
Back, Foa, & Carroll, 2001). Another study of WCDVS data showed no differential
drop out between treatment groups based on baseline posttraumatic stress symptom
level (Amaro et al., 2007). However, it must be kept in mind that previous analyses of
the Los Angeles site data indicated that women in the comparison condition with higher

121
levels of posttraumatic stress symptoms left their original treatment program earlier
than those with lower symptoms levels, while no such effect was seen in the integrated
services group (Gatz et al., 2004), an effect that was not testable using the models in
this study.
In addition to the other factors mentioned, the limited evidence of a clear effect
of trauma-focused treatment on posttraumatic stress symptoms may be related to the
underlying mechanisms maintaining PTSD. Avoidance of traumatic reminders is
believed to be a key component in the maintenance of distress and/or dysfunction
following the development of PTSD (Foa et al., 2006; Foa & Kozak, 1986; Foa &
Meadows, 1997). As a first stage trauma therapy, the trauma-focused treatment used in
this study, Seeking Safety, does not include any element of exposure to traumatic
material (Najavits, 2002). This is similar to other models of trauma-focused treatment
used in the national WCDVS; all of which focused on skill building, psychoeducation,
group discussion/ interaction, and interpersonal skills (McHugo, Kammerer et al.,
2005). As was seen in the WCDVS (Gatz et al., 2005), the women in this study
reported repeated interpersonal traumas (e.g., an average of 16 different lifetime events
and an average of four recent stressful and traumatic events in the six months prior to
the study) with most experiencing their first trauma during childhood or adolescence.
An exposure based protocol may be required to for sustained and larger changes in
posttraumatic stress symptoms to be seen among this highly and chronically
traumatized group. Cognitive distortions and negative assumptions about each

122
woman’s ability to cope (Janoff-Bulman, 1992) that support avoidance and other
symptom groups, such as numbing and hyperarousal symptoms, may require an
exposure protocol to demonstrate significant changes in posttraumatic stress symptoms.
While Seeking Safety does teach some cognitive skills (Najavits, 2002), research has
shown that cognitive restructuring did not significantly add to reductions in
posttraumatic stress symptoms when added to a prolonged exposure protocol among
women with PTSD (Foa & Rauch, 2004).
One unique aspect of this study was the measure of posttraumatic stress
symptoms. While derived from the commonly employed Posttraumatic Stress
Symptom Scale – Self Report (Foa et al., 1993), the measure used recent advances in
psychometric assessment to refine the measure so that it consistently measured
posttraumatic stress symptoms across time.. The measure used here allows for greater
confidence in the interpretation of changes in symptoms across time (Meredith & Horn,
2001). While removal of poorly performing items most strongly affected the intrusion
and avoidance subscales; leaving the scale heavily dominated by measures of numbing
and hyperarousal symptoms, results were replicated on the complete scale with no
significant differences. In addition, expected correlates, such as the relationship
between high levels of baseline unsafe events and higher levels of baseline
posttraumatic stress symptoms, were seen.
Unsafe events. The use of a behavioral measure of safety-related stressful and
traumatic events was a novel inclusion in this study. Rather than arguing that women

123
are responsible for all events of these types, which would be a kin to victim blaming in
some cases, this measure focused on events that were hypothesized to occur less
frequently if women employed safety-related coping skills consistent with the trauma-
focused treatment program (e.g., reduced substance abuse, engagement in safer
relationships, use of appropriate emotional regulation skills, psychoeducation about and
increased understanding of relationships between PTSD and substance abuse, and
increased personal responsibility). In fact, Seeking Safety presents women with more
than 80 safe coping skills and focuses session time on helping women understand their
use of poor coping choices and find other solutions (Najavits, 2003).
While the reduction of unsafe events during the first six months of the study and
the maintenance of these treatment gains among women who took more classes is
encouraging, it is not completely clear whether this reduction is due to attending
Seeking Safety or because women were in controlled residential treatment settings
where unsafe events were likely to occur. In addition, women in the comparison group
also demonstrated significant reductions in unsafe events at six-months. Previous
research on the impact of increased coping skills among individual with dual diagnoses
found that increased general coping (i.e., those helping individuals to cope with life
stressors) predicted improvement in psychological functioning (Drake, Mueser,
Brunette, & McHugo, 2004; Moggi, Ouimette, Moos, & Finney, 1999) and within the
WCDVS women who gained coping skills had significant reductions on measures of

124
general mental health symptoms, posttraumatic stress symptoms, and substance use
(Gatz et al., 2007).
It is important to note that women were able to maintain treatment gains related
to unsafe events across time. Seeking Safety and other cognitive behavioral based
protocols have been shown to successfully address posttraumatic stress symptoms and
substance abuse symptoms during the treatment phase (Desai et al., 2008; Najavits et
al., 2006; Najavits, Rosier, Nolan, & Freeman, 2007; Najavits et al., 1998; Zlotnick et
al., 2003) and some studies have shown maintenance of treatment gains up to six
months later (Desai et al., 2008; Hien et al., 2004). This is encouraging as this
population often presents with difficulties across a number of psychological and social
areas (Becker et al., 2005; P. J. Brown, Read, & Kahler, 2003), which may not be
addressed by skills learned in these programs (Cohen & Hien, 2006).
In addition to trauma-focused treatment, only one other variable predicted
significant changes in unsafe events in the final model; posttraumatic stress symptoms.
However, these effects were inconsistent across time. Higher levels of posttraumatic
stress symptoms at 3-months predicted significant reductions in unsafe events during
the last six months of the study while higher levels of posttraumatic stress symptoms at
6-months predicted significant increases in unsafe events during the last six months of
the study. Given the risk of relapse (Stewart et al., 1998) and retraumatization seen
among those with PTSD and posttraumatic stress symptoms (Koenig et al., 2004), it is
understandable that increased symptoms at the beginning of a six-month period could

125
put women at greater risk for unsafe events. However, it is unclear why symptoms
levels three months earlier predicted reductions in harmful events. Perhaps women with
lower symptom levels were less likely to continue practicing their acquired skills or
overestimated the safety of environments and people.
One potential problem with the unsafe events scale used in this study was that
potential psychological differences in how stressful or traumatic the different types of
events were was not modeled. For example, the scale included clearly different types of
events, such as serious money problems, which while stressful, was probably
qualitatively different than events such as physical and sexual assault. However, each
contributed equally to the summed scale score. One way future studies of the effects of
trauma-focused treatment on unsafe events could move the field forward is by
developing a measure of severity of threat for each type of unsafe event. One model for
assessing threat was developed by Brown (1989) and described in detail by (Kendler &
Prescott, 2006) defines threat as the degree to which the events require people to
reassess central aspects of their self-identity or life plans and is assessed by trained
interviewers. Specific coping skills may be better suited for helping individuals avoid
higher threat events and could be prioritized in treatment.
Other statistical methods can be used to assess if how long trauma-focused
treatment keeps unsafe events from occurring (e.g., survival models) or research can
focus on treatment reducing the probability of specific events occurring (e.g.,
probability transition modeling). However, these require interesting decisions to be

126
made about revictimization and treatment; are we seeking to completely eliminate
unsafe events, lessen their probability, and, if so, what is an acceptable level of unsafe
events?
Time in residential treatment. The time each woman spent engaged in
residential treatment was included as a control variable to help account for the effects of
treatment, in general, for possible increases in posttraumatic stress symptoms following
earlier treatment engagement (Stewart et al., 1999), and the impact of being in
controlled environments on possible reductions in unsafe events. The most consistent
effect of trauma-focused treatment, seen across the length of the study, showed that
greater engagement in the program predicted increased use of residential treatment
services. During the first six months of the study, women had to be engaged in
residential treatment to attend Seeking Safety; however, most women completed the
trauma-focused classes by six months, and, thus, continued to use residential services
after having done so. One limitation is that the variable does not represent engagement
in a single program, but combines women’s report on participation in any residential
services. This effect of longer engagement is hopeful, because previous research
suggests that as many as 50% of those entering substance abuse treatment drop out
before the first 30 days (Ashley, Marsden, & Brady, 2003; Stark, 1992) and that longer
treatment is related to better treatment outcomes on some mental health outcomes
(Simpson et al., 1997), including results from the WCDVS (Amaro et al., 2007), though
significant differences of treatment length on posttraumatic stress symptoms were not

127
found (Amaro et al., 2007). However, this effect must be evaluated in light of the
differences in treatment program lengths between WCDVS treatment groups.
Integrated treatment programs were longer than comparison sites, though comparison
often referred women to controlled environments following treatment completion,
which may have lessen the impact of program differences for days in residential
treatment.
Previous research has shown that coping strategies are related to retention in
substance abuse treatment (Kohn, Mertens, & Weisner, 2002). Results using WCDVS
data from the Los Angeles site (Gatz et al., 2004) indicated that women who received
integrated treatment services, especially those who completed their residential treatment
program, showed significant gains in coping skills similar to those learned in Seeking
Safety, while women in the comparison group showed less coping skill use at the end of
the study. Also, researchers found that treatment gains at 12-months were partially
mediated by gains in coping skills (Gatz et al., 2007). However, it is also important to
note that women who completed treatment programs were less distressed at baseline
and at 12-months compared to women who did not complete their treatment programs
(Gatz et al., 2007). Future research needs to investigate the impact of acquisition and
practice of coping skills on long-term treatment gains.
Limitations. The results need to be interpreted in light of several study
limitations. First, women participating in the WCDVS were recruited from the
treatment programs they chose to attend or to which they were legally mandated to

128
attend, and, therefore, there was no random assignment of women to WCDVS treatment
arm. Thus, the possible detrimental effects of non-equivalent control groups must be
considered. Partially mitigating this concern is that women from the Los Angeles site
did not significant differ at baseline on a broad range of demographic, psychosocial, and
mental health variables, including posttraumatic stress symptoms and coping skills
(Gatz et al., 2007) and women in this study (i.e. a small number of women who were
not in residential services from the Los Angeles site data were dropped in this study)
did not differ significantly on baseline levels of posttraumatic stress symptoms, unsafe
events or time spent in residential treatment during the first three months of the study.
However, there were some important baseline differences between treatment
groups with women receiving integrated services reporting greater rates of lifetime and
recent incarceration as well as greater likelihood of being mandated for treatment (Gatz
et al., 2007; Gatz et al., 2004). Research indicates that women who were referred
through criminal justice system (Ashley et al., 2003) or mandated to treatment (Brecht,
Greenwell, & Anglin, 2005) were more likely to complete program their programs as
well as remain in treatment longer (Amaro et al., 2007). Also, in this study women in
the comparison group reported significantly longer periods of residential treatment in
the three months prior to the study compared to women in the integrated treatment
group. This is contrasted with women in the integrated treatment group spending
significantly more days in jail in the three months prior to the study compared to
women in the comparison group. While the impact of residential treatment was

129
adjusted for statistically, the differences in incarceration variables and mandated
treatment were not accounted for. Recent research on WCDVS data from the Los
Angeles and Boston sites indicated that when baseline differences were accounted for,
women receiving integrated treatment services at a 31% lower risk of dropout from
treatment compared to women in the comparison group (Amaro et al., 2007). In
addition, previously discussed differences in program lengths also need to be
considered. However, overall, it is key to remember that the women in the two
treatment arms were remarkably similar at baseline.
Another concern is the effects of differential attrition from both in the larger
study. Previous analyses of the Los Angeles site data indicated that women at
comparison group with higher initial posttraumatic stress symptoms scores were less
likely to be interviewed at 12-months compared to women who were re-assessed at 12-
months, but no such effect was found for integrated treatment group (Gatz et al., 2004).
While posttraumatic stress symptom levels did predict unsafe events, the effect was not
consistent across time, and the differential attrition is not likely to be sole reason for
changes in unsafe events seen in this study.
While this study attempted to isolate the effects of the trauma-focused treatment,
the larger WCDVS was not designed to specifically assess the effects of this program,
which was tightly bundled into the broader integrated service package. Thus,
conclusions, especially in regards to posttraumatic stress symptoms, need to be
replicated in further studies. Where possible, advanced statistical techniques were used

130
to partially address this concern and the moderating effect of trauma-focused treatment
on unsafe events was clearly seen, independent of the similar changes in the comparison
group. This study also used a novel behavioral trauma measure theoretically expected
to vary with level of participation in the trauma-focused treatment group, as women
were taught and practiced safety behaviors (Najavits, 2003). Despite these
considerations, the results of this study are based on a study with high external validity,
a large sample size for this type of population, good study retention rates (Gatz et al.,
2004), few significant baseline differences in groups (Gatz et al., 2007), and advanced
statistical procedures used to isolate relevant effects.
Finally, while advanced statistical models were used to try to disentangle effects
of trauma-focused treatment from other treatment services, the data likely violated
assumptions underlying the use of these models. For example, there was high levels of
variability within the longitudinal variables, especially residential treatment, indicating
presence of subpopulations, but we did not identify factors accounting for the variability
in data. With the residential treatment variable, we allowed residuals to be correlated
across time in order to model the data. This heterogeneity also may have masked any
significant effects of trauma-focused on outcome measures. There was more than one
model of latent change over time present and most models required non-linear change
effects had to be included. Developing more detailed understanding of the course of
these variables across time and factors influencing change should be conducted before
including in more complex models in the future.

131
Implications for Clinical Practice. The recent and expanding call for and
research on providing comprehensive, integrated, trauma-informed, and consumer-
involved services to individuals with co-occurring disorders represents an important
step forward in the treatment of this vulnerable population (Alexander, 1996; Elliott et
al., 2005; Mockus et al., 2005; RachBeisel et al., 1999; Watkins et al., 2001). Results
showed that trauma-focused treatment, like that in Seeking Safety, demonstrated some
important benefits for women with co-occurring mental health and substance abuse
disorders. While research has shown that this can be a difficult population to treat as
they have significant concerns across multiple domains (Becker et al., 2005; Brunette,
Mueser, & Drake, 2004; Drake et al., 2008), this study demonstrated high use of
residential services after an initial level of engagement, and reductions in traumatic and
highly stressful events likely promoted by increased use of safety behaviors. Previous
studies using the WCDVS’ Los Angeles data indicated that women found the treatment
program highly relevant to their needs (V. B. Brown et al., 2007; Gatz et al., 2007) and
demonstrated increased coping skills (Gatz et al., 2007). While Seeking Safety
attendance was not related to significant changes in posttraumatic stress symptoms in
this study and previous research found women still reported moderate levels of
posttraumatic stress symptoms at the conclusion of the study (Gatz et al., 2004), women
appeared to develop relevant first stage skills to prepare them for future exposure-based
protocols.

132
In sum, in this study of women with co-occurring mental health and substance
abuse disorders and histories of interpersonal violence, those who participated in
trauma-focused treatment demonstrated significant reductions in unsafe behaviors that
were maintained across time as well as increased participation in residential treatment.
Women in a comparison group also showed significant reductions in unsafe events.
Thus, while women with co-occurring disorders can be a difficult population to engage
in treatment and who experience many psychosocial burdens (Becker et al., 2005;
Drake et al., 2008), adding a trauma-focused treatment component to integrated mental
health and substance abuse service programs was found positively to affect women’s
safety behaviors and engagement in residential treatment.

133
References

Abueg, F. R., & Fairbank, J. A. (1992). Behavioral treatment of posttraumatic stress
disorder and co-occurring substance abuse: A multidimensional stage model. In
P. A. Saigh (Ed.), Posttraumatic stress disorder: A behavioral approach to
assessment and treatment (pp. 111-146). Boston: Allyn & Bacon.
Akaike, H. (1973). Information theory and an extension of the maximum likelihood
principle. In B. N. Petrov & F. Csaki (Eds.), Second International Symposium on
Information Theory (pp. 267-281). Budapest: Akademiai Kiado.
Alexander, M. J. (1996). Women with co-occurring addictive and mental disorders: An
emerging profile of vulnerability. American Journal of Orthopsychiatry, 66, 61-
70.
Amaro, H., Chernoff, M., Brown, V., Arévalo, S., & Gatz, M. (2007). Does integrated
trauma-informed substance abuse treatment increase treatment retention?
Journal of Community Psychology, 35, 845-862.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental
disorders (4th ed.). Washington D.C.: American Psychiatric Association.
Andrews, L., Joseph, S., Shevlin, M., & Troop, N. (2006). Confirmatory factor analysis
of posttraumatic stress symptoms in emergency personnel: An examination of
seven alternative models. Personality and Individual Differences, 41, 213-224.
Ashley, O. S., Marsden, M. E., & Brady, T. M. (2003). Effectiveness of Substance
Abuse Treatment Programming For Women: A Review. The American Journal
of Drug and Alcohol Abuse, 29(1), 19-53.
Asmundson, G. J. G., Frombach, I., McQuaid, J., Pedrelli, P., Lenox, R., & Stein, M. B.
(2000). Dimensionality of posttraumatic stress symptoms: A confirmatory
factor analysis of DSM-IV symptom clusters and other symptom models.
Behaviour, Research and Therapy, 38, 203-214.
Asmundson, G. J. G., Wright, K. D., McCreary, D. R., & Pedlar, D. (2003). Post-
traumatic stress disorder symptoms in United Nations peacekeepers: An
examination of factor structure in peacekeepers with and without chronic pain.
Cognitive Behaviour Therapy, 32, 26-37.

134
Baschnagel, J. S., O'Connor, R. M., Colder, C. R., & Hawk, L. W. (2005). Factor
structure of posttraumatic stress among western New York undergraduates
following the September 11th terrorist attack on the World Trade Center.
Journal of Traumatic Stress, 18, 677-684.
Becker, M. A., Noether, C. D., Larson, M. J., Gatz, M., Brown, V. B., Heckman, J. P.,
et al. (2005). Characteristics of women engaged in treatment for trauma and co-
occurring disorders: Findings from a national multisite study. Journal of
Community Psychology, 33, 429-443.
Bentler, P. M. (1990a). Comparative fix indexes in structural models. Psychological
Bulletin, 107, 238-246.
Bentler, P. M. (1990b). EQS: A structural equation program. Los Angeles: BMDP
Statistical Software.
Bradley, R. G., & Follingstad, D. R. (2003). Group therapy for incarcerated women
who experienced interpersonal violence: A pilot study. Journal of Traumatic
Stress, 16, 337-340.
Brady, K. T., Dansky, B. S., Back, S. E., Foa, E. B., & Carroll, K. M. (2001). Exposure
therapy in the treatment of PTSD among cocaine-dependent individuals:
Preliminary findings. Journal of Substance Abuse Treatment, 21, 47-54.
Brecht, M.-L., Greenwell, L., & Anglin, M. D. (2005). Methamphetamine treatment:
Trends and predictors of retention and completion in a large state treatment
system (1992-2002). Journal of Substance Abuse Treatment, 29, 295-306.
Brewin, C. R., Dalgleish, T., & Joseph, S. (1996). A dual representation theory of
posttraumatic stress disorder. Psychological Review, 103, 670-686.
Briere, J., Kaltman, S., & Green, B. L. (2008). Accumulated childhood trauma and
symptom complexity. Journal of Traumatic Stress, 21, 223-226.
Briere, J., & Runtz, M. (1987). Post sexual abuse trauma: Data implications for clinical
practice. Journal of Interpersonal Violence, 2, 367-379.
Briere, J., Woo, R., McRae, B., Foltz, J., & Sitzman, R. (1997). Lifetime victimization
history, demographics, and clinical status in female psychiatric emergency room
patients. Journal of Nervous and Mental Disease, 185, 95-101.

135
Brown, G. W. (1989). Life events and illness. London: University of London.
Brown, P. J. (2000). Outcome in female patients with both substance use and post-
traumatic stress disorders. Alcoholism Treatment Quarterly, 18, 127-135.
Brown, P. J., Read, J. P., & Kahler, C. W. (2003). Comorbid posttraumatic stress
disorder and substance use disorders: Treatment outcomes and the role of
coping. In P. Ouimette & P. J. Brown (Eds.), Trauma and Substance Abuse:
Causes, Consequences and Treatment of Comorbid Disorders (pp. 171 - 188).
Washington, DC: American Psychological Association.
Brown, P. J., Recupero, P. R., & Stout, R. L. (1995). PTSD substance abuse
comorbidity and treatment utilization. Addictive Behaviors, 20, 251-254.
Brown, V. B., Huba, G. J., & Melchior, L. A. (1995). Level of burden: Women with
more than one co-occurring disorder. Journal of Psychoactive Drugs, 27, 339-
346.
Brown, V. B., Najavits, L. M., Cadiz, S., Finkelstein, N., Heckman, J. P., &
Rechberger, E. (2007). Implementing an evidence-based practice: Seeking
Safety group. Journal of Psychoactive Drugs, 39, 231 - 240.
Brunette, M., Mueser, K. T., & Drake, R. E. (2004). A review of research on residential
programs for people with severe mental illness and co-occurring substance use
disorders. Drug and Alcohol Review, 23, 471 - 481.
Buckley, T. C., Blanchard, E. B., & Hickling, E. J. (1996). A prospective examination
of delayed onset PTSD secondary to motor vehicle accidents. Journal of
Abnormal Psychology, 105, 617-625.
Buckley, T. C., Blanchard, E. B., & Hickling, E. J. (1998). A confirmatory factor
analysis of posttraumatic stress symptoms. Behavior Research and Therapy, 36,
1091-1099.
Chilcoat, H. D., & Menard, C. (2003). Epidemiological investigations: Comorbidity of
posttraumatic stress disorder and substance use disorder. In P. C. Ouimette & P.
J. Brown (Eds.), Trauma and Substance Abuse: Causes, consequences, and
treatment of comorbid disorders (pp. 9-28). Washington, DC: American
Psychological Association.
Chung, S., Domino, M. E., Jackson, E. W., & Morrissey, J. (2007). Reliability of self-
reported health service use: Evidence from the Women with Co-occurring
Disorders, and Violence Study. Journal of Behavioral Health Services and
Research, 35, 265 - 278.

136
Classen, C. C., Palesh, O. G., & Aggarwal, R. (2005). Sexual revictimization: A review
of the empirical literature. Trauma Violence Abuse, 6, 103-129.
Clay, K. M., Olsheski, J. A., & Clay, S. W. (2000). Alcohol use disorders in female
survivors of childhood sexual abuse. Alcoholism Treatment Quarterly, 18, 19-
29.
Cloitre, M., Cohen, L. R., Edelman, R. E., & Han, H. (2001). Posttraumatic Stress
Disorder and extent of trauma exposure as correlates of medical problems and
perceived health among women with childhood abuse. Women and Health, 34,
1-18.
Cloitre, M., Koenen, K. C., Cohen, L. R., & Han, H. (2002). Skills training in affective
and interpersonal regulation followed by exposure: A phase-based treatment for
PTSD related to childhood abuse. Journal of Consulting and Clinical
Psychology, 70, 1067-1074.
Cloitre, M., Scarvalone, P., & Difede, J. (1997). Posttraumatic Stress Disorder, self- and
Interpersonal dysfunction among sexually retraumatized women. Journal of
Traumatic Stress, 10, 437-452.
Cocozza, J., Jackson, E. W., Hennigan, K., Morrissey, J. P., Glover Reed, B., Fallot, R.,
et al. (2005). Outcomes for women with co-occurring disorders and trauma:
Program-level effects. Journal of Substance Abuse Treatment, 28, 109-119.
Coffey, S. F., Dansky, B. S., Falsetti, S. A., Saladin, M. E., & Brady, K. T. (1998).
Screening for PTSD in a substance abuse sample: Psychometric properties of a
modified version of the PTSD Symptom Scale Self-Report. Journal of
Traumatic Stress, 11, 393-399.
Cohen, L. R., & Hien, D. A. (2006). Treatment outcomes for women with substance
abuse and PTSD who have experienced complex trauma. Psychiatric Services,
57, 100-106.
Coid, J., Peruckevitch, A., Feder, G., Chung, W.-S., Richardson, J., & Moorey, S.
(2001). Relation between childhood sexual and physical abuse and risk or
revicitimisation in women: A cross-sectional survey. Lancet, 358, 450-454.
Coker, A. L., Smith, P. H., McKeown, R. E., & King, M. J. (2000). Partner violence by
type: Physical, sexual, and psychological battering. American Journal of Public
Health, 90, 553-559.
Comfort, M., & Kaltenbach, K. (2000). Predictors of treatment outcomes for substance-
abusing women: A retrospective study. Substance Abuse, 21, 33-45.

137
Cook, J. M., Walser, R. D., Kane, V., Ruzek, J. I., & Woody, G. (2006). Dissemination
of feasibility of a cognitive-behavioral treatment for substance abuse use
disorders and posttraumatic stress disorder in the Veterans Administration.
Journal of Psychoactive Drugs, 38, 89-92.
Cordova, M. J., Studts, J. L., Hann, D. M., Jacobsen, P. B., & Andrykowski, M. A.
(2000). Symptoms structure of PTSD following breast cancer. Journal of
Traumatic Stress, 13, 301-319.
Courtois, C. A. (1988). Healing the Wounds of Incest. New York: W.W. Norton.
Cunningham, J., Pearce, T., & Pearce, P. (1988). Childhood sexual abuse and medical
complaints in adult women. Journal of Interpersonal Violence, 3, 131-144.
Cusack, K., Morrissey, J., & Ellis, A. R. (2008). Targeting trauma-related interventions
and improving outcomes for women with co-occurring disorders. Administration
and Policy in Mental Health and Mental Health Services Research, 35, 147-158.
Derogatis, L. R. (1993). Brief symptom inventory: Administration, scoring and
procedures manual: Fourth Edition (Fourth ed.). Minneapolis, MN: NCS
Pearson, Inc.
Desai, R. A., Harpaz-Rotem, I., Najavits, L. M., & Rosenheck, R. A. (2008). Impact of
Seeking Safety program on clinical outcomes among homeless female veterans
with psychiatric disorders. Psychiatric Services, 59, 996-1003.
Drake, R. E., Mueser, K. T., Brunette, M. F., & McHugo, G. J. (2004). A review of
treatments for people with severe mental illnesses and co-occurring substance
use disorders. Psychiatric Rehabilitation Journal, 27, 360-374.
Drake, R. E., O'Neal, E. L., & Wallach, M. A. (2008). A systematic review of
psychosocial research on psychosocial interventions for people with co-
occurring severe mental and substance use disorders. Journal of Substance
Abuse Treatment, 34, 123-138.
DuHamel, K. N., Ostroff, J., Ashman, T., Winkel, G., Mundy, E. A., Keane, T. M., et
al. (2004). Construct validity of the Posttraumatic Stress Disorder Checklist in
cancer survivors: Analyses based on two samples. Psychological Assessment,
16, 255-266.
Duncan, T. E., Duncan, S. C., & Strycker, L. A. (2006). An Introduction to Latent
Variable Growth Curve Modeling. Mahwa, NJ: Lawrence Erlbaum.

138
Duncan, T. E., Duncan, S. C., Strycker, L. A., & Li, F. (2002). A latent variable
framework for power estimation within intervention contexts. Journal of
Psychopathology and Behavioral Assessment, 24, 1-12.
El-Bassel, N., Gilbert, L., Wu, E., Go, H., & Hill, J. (2005). Relationship between drug
abuse and intimate partner violence: A longitudinal study among women
receiving methadone. American Journal of Public Health, 96, 465-470.
Elliott, D. E., Bjelajac, P., Fallot, R. D., Markoff, L. S., & Reed, B. G. (2005). Trauma-
informed or trauma-denied: Principles and implementation of trauma-informed
services for women. Journal of Community Psychology, 33, 461-477.
Ellis, A. R., & Morrissey, J. (2009). Assessing multiple outcomes of women with co-
occurring disorders and trauma in a multi-site trail: A propensity score
approach. Administration and Policy in Mental Health and Mental Health
Services Research, 35, 147 - 158.
Fan, X. (2003). Power of latent growth modeling for detecting group differences in
linear growth trajectory parameters. Structural Equation Modeling, 10, 380-400.
Feuer, C. A., Nishith, P., & Resick, P. A. (2005). Prediction of numbing and effortful
avoidance in female rape survivors with chronic PTSD. Journal of Traumatic
Stress, 18, 165-170.
Flack, W. F., Litz, B. T., Hsieh, F. Y., & Kaloupek, D. G. (2000). Predictors of
emotional numbing, revisited: A replication and extension. Journal of
Traumatic Stress, 13, 611-618.
Foa, E. B., Huppert, J. D., & Cahill, S. P. (2006). Emotional processing theory: An
update. In B. O. Rothbaum (Ed.), Pathological anxiety: Emotional processing
in etiology and treatment (pp. 3-24). New York: Guilford.
Foa, E. B., & Kozak, M. J. (1986). Emotional processing of fear: Exposure to
corrective information. Psychological Bulletin, 99, 20-35.
Foa, E. B., & Meadows, E. A. (1997). Psychosocial treatments for posttraumatic stress
disorder: A critical review. Annual Review of Psychology, 48(449-480).
Foa, E. B., & Rauch, S. A. M. (2004). Cognitive changes during prolonged exposure
versus prolonged exposure plus cognitive restructuring in female assault
survivors with posttraumatic stress disorder. Journal of Consulting and Clinical
Psychology, 72, 879-884.

139
Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993). Reliability and
validity of a brief instrument for assessing post-traumatic stress disorder.
Journal of Traumatic Stress, 6, 459-473.
Foa, E. B., Riggs, D. S., & Gershuny, B. S. (1995). Arousal, numbing, and intrusion:
Symptom structure of PTSD following assault. American Journal of Psychiatry,
152, 116-120.
Follette, V. M., Polusny, M. A., Bechtle, A. E., & Naugle, A. E. (1996). Cumulative
trauma: The impact of child sexual abuse, adult sexual assault, and spouse
abuse. Journal of Traumatic Stress, 9, 25-35.
Fund, C. (1996). Violence against women in the United States: A comprehensive
background paper. New York: Commonwealth Fund.
Gatz, M., Brown, V., Hennigan, K., Rechberger, E., O'Keefe, M., Rose, T., et al.
(2007). Effectiveness of an integrated trauma-informed approach to treating
women with co-occurring disorders and histories of trauma. Journal of
Community Psychology, 35, 863-878.
Gatz, M., Hennigan, K., O'Keefe, M., & Rose, T. (2004). Women, co-occurring
disorders, and violence study: Evaluation report. Culver City: PROTYPES,
Centers for Innovation in Health, Mental Health, and Social Services.
Gatz, M., Russell, L. A., Grady, L. A., Kram-Fernandez, D., Clark, C., & Marshall, B.
(2005). Women's recollections of victimization, psychological problems, and
substance use. Journal of Community Psychology, 33, 479-493.
Gerstein, D. R., Datta, A. R., Ingels, J. S., Johnson, R. A., Rasinski, K. A., Schildhaus,
S., et al. (1997). NTIES. National Treatment Improvement Evaluation Study.
Final Report. Rockville, MD: Center for Substance Abuse Treatment, Substance
Abuse and Mental Health Services Administration.
Giard, J., Hennigan, K., Huntington, N., Vogel, W., Rinehart, D., Mazelis, R., et al.
(2005). Development and implementation of a multisite evaluation for the
Women, Co-Occurring Disorders and Violence Study. Journal of Community
Psychology, 33, 411-427.
Gold, S. R., Milan, L. D., Mayall, A., & Johnson, A. E. (1994). A cross-validation study
of the Trauma Symptom Checklist: The role of mediating variables. Journal of
Interpersonal Violence, 9, 12-26.

140
Golding, J. M. (1999). Sexual-assault history and long-term physical health problems:
Evidence from clinical and population epidemiology. Current Directions in
Psychological Science, 8, 191-194.
Gorcey, M., Santiago, J. M., & McCall-Perez, F. (1986). Psychological consequences
for women sexually abused in childhood. Social Psychiatry and Psychiatric
Epidemiology, 21, 129-133.
Greenfield, S. F., Brooks, A. J., Gordon, S. M., Green, C. A., Kropp, F., McHugh, R.
K., et al. (2007). Substance abuse treatment entry, retention, and outcome in
women: A review of the literature. Drug and Alcohol Dependence, 86(1), 1-21.
Grella, C. E. (1996). Background and overview of mental health and substance abuse
treatment systems: Meeting the needs of women who are pregnant or parenting.
Journal of Psychoactive Drugs, 28, 319-343.
Griesel, D., Wessa, M., & Flor, H. (2006). Psychometric qualities of the German
Version of the Posttraumatic Diagnostic scale (PTDS). Psychological
Assessment, 18, 262-268.
Harris, M. (1994). Modifications in service delivery and clinical treatment for women
diagnosed with severe mental illness who are also survivors of trauma. Journal
of Mental Health Administration, 21, 397-406.
Harris, M., & Fallot, R. (2001). Using trauma theory to design service systems. San
Francisco: Jossey-Bass.
Hendrickson, E. L., Schmal, M. S., & Ekleberry, S. C. (2004). Treating Co-occurring
Disorders: A Handbook for Mental Health and Substance Abuse Professionals.
Binghamton, NY: Haworth Press.
Herman, J. (1992). Trauma and Recovery: The Aftermath of Violence - From Domestic
Abuse to Political Terror. New York: Basic books.
Hertzog, C., von Oertzen, T., Ghisletta, P., & Lindenberger, U. (2008). Evaluating the
power of latent growth curve models to detect individual differences in change.
Structural Equation Modeling, 15, 541-563.
Hien, D. A., Cohen, L. R., Miele, G. M., Litt, L. C., & Capstick, C. (2004). Promising
treatments for women with comorbid PTSD and substance use disorders.
American Journal of Psychiatry, 161, 1426-1432.

141
Holdcraft, L. C., & Comtois, K. A. (2002). Description of and preliminary data from a
women's dual diagnosis community mental health program. Canadian Journal
of Community Mental Health, 21(91-109).
Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement
invariance in aging research. Experimental Aging Research, 18, 117-144.
Horowitz, M. J. (1986). Stress Response Syndromes. Northvale, NJ: Aronson.
Houskamp, B. M., & Foy, D. W. (1991). The assessment of Posttraumatic Stress
Disorder in battered women. Journal of Interpersonal Violence, 6, 367-375.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation
Modeling, 6, 1-55.
Huntington, N., Jahn Moses, D., & Veysey, B. M. (2005). Developing and
implementing a comprehensive approach to serving women with co-occurring
disorders and histories of trauma. Journal of Community Psychology, 33, 395-
410.
Jacobsen, P. B., Southwick, S. M., & Kosten, T. R. (2001). Substance use disorders in
patients with posttraumatic stress disorder: A review of the literature. American
Journal of Psychiatry, 158, 1184-1190.
Janoff-Bulman, R. (1992). Shattered assumptions: Towards a new psychology of
trauma. New York: Free Press.
Kendler, K. S., & Prescott, C. A. (2006). Genes, environment, and psychopathology:
Understanding the causes of psychiatric and substance use disorders. New
York: Guilford Press.
Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., Nelson, C. B., & Breslau, N.
(1999). Epidemiologic risk factors for trauma and PTSD. In R. Yehuda (Ed.),
Risk factors for Posttraumatic Stress Disorder (pp. 23 - 59). Washington, DC:
American Psychiatric Press.
Kilpatrick, D. G., Acierno, R., Resnick, H. S., Saunders, B. E., & Best, C. L. (1997). A
2-year longitudinal analysis of the relationships between violent assault and
substance use in women. Journal of Consulting and Clinical Psychology, 65,
834-847.

142
King, D. W., King, L. A., Fairbank, J. A., Schlenger, W. E., & Surface, R. C. (1993).
Enhancing the precision of the Mississippi Scale for Combat-Related
Posttraumatic Stress Disorder: An application of Item Response Theory.
Psychological Assessment, 5, 457-471.
King, D. W., Leskin, G. A., King, L. A., & Weathers, F. W. (1998). Confirmatory
factor analysis of the Clinician-Administered PTSD scale: Evidence for the
dimensionality of posttraumatic stress disorder. Psychological Assessment, 10,
90-96.
King, L. A., & King, D. W. (1994). Latent structure of the Mississippi Scale for
Combat-Related Post-traumatic Stress Disorder: Exploratory and Higher-Order
Confirmatory Factor Analysis. Assessment, 1, 275-291.
King, L. A., King, D. W., Fairbank, J. A., Keane, T. M., & Adams, G. A. (1998).
Resilience-recovery factors in post-traumatic stress disorder among female and
male Vietnam veterans: Hardiness, postwar social support, and additional
stressful life events. Journal of Personality and Social Psychology, 74, 420-434.
King, L. A., King, D. W., McArdle, J. J., Saxe, G. N., Doron-LaMarca, S., & Orazem,
R. J. (2006). Latent difference score approach to longitudinal trauma research.
Journal of Traumatic Stress, 19, 771-785.
Kirmayer, L. J., Lemelson, R., & Barad, M. (2007). Understanding Trauma:
Integrating Biological, Clinical and Cultural Perspectives. Cambridge:
Cambridge University Press.
Koenig, L. J., Doll, L. S., O'Leary, A., & Pequegnat, W. (2004). From Child Sexual
Abuse to Adult Sexual Risk: Trauma, Revictimization and Intervention.
Washington, DC: American Psychological Association.
Kohn, C. S., Mertens, J. R., & Weisner, C. M. (2002). Coping among individuals
seeking private chemical dependence treatment: Gender differences and impact
on length of stay in treatment. Alcoholism: Clinical and Experimental
Research, 26, 1228-1233.
Larson, M. J., Miller, L., Becker, M. A., Richardson, E., Kammerer, N., Thom, J., et al.
(2005). Physical health burdens of women with histories of co-occurring
substance abuse and mental disorders. Journal of Behavioral Health Services
and Research, 32, 128 - 140.
Linehan, M. M. (1993). Cognitive-behavioral treatment of borderline personality
disorder. New York: Guilford Press.

143
Linehan, M. M., Bohus, M., & Lynch, T. R. (2007). Dialectical behavior therapy for
pervasive emotional dysregulation. In J. J. Gross (Ed.), Handbook of Emotion
Regulation (pp. 581 - 606). New York Guilford.
Litz, B. T., Schlenger, W. E., Weathers, F. W., Caddell, J. M., Fairbank, J. A., &
LaVange, L. M. (1997). Predictors of emotional numbing in posttraumatic stress
disorder. Journal of Traumatic Stress, 10, 607-618.
MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and
determination of sample size for covariance structure modeling. Psychological
Methods, 1, 130-149.
Marshall, G. N. (2004). Posttraumatic Stress Disorder Symptom Checklist: Factor
structure and English-Spanish measurement invariance. Journal of Traumatic
Stress, 17, 223-230.
McArdle, J. J., Fisher, G. G., & Kadlec, K. M. (in press). Latent variable analyses of
age trends in cognition from the Health and Retirement Study, 1992 - 2004.
Psychology and Aging.
McArdle, J. J., & Hamagami, F. (2001). Latent difference score structural models for
linear dynamic analyses with incomplete longitudinal data. In L. M. Collins &
A. G. Sayer (Eds.), New Methods for the Analysis of Change (pp. 105-136).
Washington, DC: American Psychological Association.
McArdle, J. J., & Nesselroade, J. R. (1994). Using multivariate data to structure
developmental change. In S. H. Cohen & H. W. Reese (Eds.), Life-span
Developmental Psychology: Methodological Innovations (pp. 223-267).
Hillsdale, NJ: Erlbaum.
McFarlane, A. C. (1988). The longitudinal course of posttraumatic morbidity: The
range of outcomes and their predictors. Journal of Nervous and Mental Disease,
176, 30-39.
McHugo, G. J., Caspi, Y., Kammerer, N., Mazelis, R., Jackson, E. W., Russ, E., et al.
(2005). The assessment of trauma history in women with co-occurring substance
abuse and mental disorders and a history of interpersonal violence. Journal of
Behavioral Health Services and Research, 32, 113-127.
McHugo, G. J., Kammerer, N., Jackson, E. W., Markoff, L. S., Gatz, M., Larson, M. J.,
et al. (2005). Women, co-occurring disorders, and violence study: Evaluation
design and study population. Journal of Substance Abuse Treatment, 28, 91-107.

144
McLellan, A. T., Kushner, H., Jetzger, D., Peters, F., Smith, I., Grissom, G., et al.
(1992). The fifth edition of the Addiction Severity Index: Historical critique and
normative data. Journal of Substance Abuse and Treatment, 9, 199-213.
McWilliams, L. A., Cox, B. J., & Asmundson, G. J. G. (2005). Symptom structure of
posttraumatic stress disorder in a nationally representative sample. Anxiety
Disorders, 19, 626-641.
Melchior, L. A., Huba, G. J., Brown, V. B., & Slaughter, R. (1999). Evaluation of the
effects of outreach to women with multiple vulnerabilities on entry into
substance abuse treatment. Evaluation and Program Planning, 22, 269-277.
Meredith, W., & Horn, J. L. (2001). The role of factorial invariance in modeling growth
and change. In L. M. Collins & A. Sayer (Eds.), New methods for the analysis of
change (pp. 201-240). Washington, DC: American Psychological Association.
Messman-Moore, T. L., & Long, P. J. (2002). Alcohol and substance use disorders as
predictors of child to adult sexual revictimization in a sample of community
women. Violence and Victims, 17, 319-340.
Mockus, S., Mars, L. C., Ovard, D. G., Mazelis, R., Bjelajac, P., Grady, J., et al. (2005).
Developing consumer/survivor/recovering research: Our experience with the
SAMHSA Women. Journal of Community Psychology, 33, 513-525.
Moggi, F., Ouimette, P., Moos, R., & Finney, J. W. (1999). Dual diagnosis patients in
substance abuse treatment: relationship of general coping and substance-specific
coping to 1-year outcomes. Addiction, 94, 1805-1816.
Mol, S. S. L., Arntz, A., Metsemakers, J. F. M., Dinant, G.-J., Vilters-Van Montfort, P.
A. P., & Knottnerus, J. A. (2005). Symptoms of post-traumatic stress disorder
after non-traumatic events: Evidence from an open population study. British
Journal of Psychiatry, 186, 494-499.
Morrissey, J. P., Ellis, A. R., Gatz, M., Amaro, H., Glover Reed, B., Savage, A., et al.
(2005). Outcomes for women with co-occurring disorders and trauma: Program
and person-level effects. Journal of Substance Abuse Treatment, 28, 121-133.
Morrissey, J. P., Jackson, E. W., Ellis, A. R., Amaro, H., Brown, V. B., & Najavits, L.
M. (2005). Twelve-month outcomes of trauma-informed interventions for
women with co-occurring disorders. Psychiatric Services, 56, 1213-1222.
Mueser, K. T., Noordsy, D. L., Drake, R. E., & Fox, L. (2003). Integrated Treatment for
Dual Disorders: A guide to Effective Practice. New York: Guilford Press.

145
Muthén, L. K., & Muthén, B. O. (2007). Mplus 4.2. Los Angeles: StatModel.
Najavits, L. M. (2002). Seeking safety: A treatment manual for PTSD and substance
abuse. New York: Guilford Press.
Najavits, L. M. (2003). Seeking Safety: A new psychotherapy for posttraumatic stress
disorder and substance abuse disorders. In P. Ouimette & P. J. Brown (Eds.),
Trauma and Substance Abuse: Causes, Consequences and Treatments of
Comorbid Disorders (pp. 147-170). Washington, DC: American Psychological
Association.
Najavits, L. M. (2007). Seeking Safety: An evidence-based model for substance abuse
and trauma/PTSD. In K. A. Witkiewitz & G. A. Marlatt (Eds.), Therapist's
guide to evidence based relapse prevention: Practical resources for the mental
health professional (pp. 141-167). San Diego: Elsevier Press.
Najavits, L. M., Gallop, R. J., & Weiss, R. D. (2006). Seeking Safety therapy for
adolescent girls with PTSD and substance abuse: A randomized controlled trial.
Journal of Behavioral Health Services and Research, 33, 453-463.
Najavits, L. M., Rosier, M., Nolan, A. L., & Freeman, M. C. (2007). A new gender-
based model for women's recovery from substance abuse: Results of a pilot
outcome study. American Journal of Drug and Alcohol Abuse, 33(5-11).
Najavits, L. M., Schmitz, S., Gotthardt, S., & Weiss, R. D. (2005). Seeking Safety plus
exposure therapy for dual diagnosis men. Journal of Psychoactive Drugs, 27,
425-435.
Najavits, L. M., Weiss, R. D., Shaw, S. R., & Muenz, L. R. (1998). "Seeking Safety":
Outcome of a new cognitive-behavioral psychotherapy for women with
posttraumatic stress disorder and substance dependence. Journal of Traumatic
Stress, 11, 437-456.
Palmieri, P. A., & Fitzgerald, L. F. (2005). Confirmatory factor analysis of
posttraumatic stress symptoms in sexually harassed women. Journal of
Traumatic Stress, 18, 657-666.
Port, C. L., Engdahl, B., & Frazier, P. (2001). A longitudinal and retrospective study of
PTSD among older prisoners of war. American Journal of Psychiatry, 158,
1474-1479.
Preacher, K. J., Wichman, A. L., MacCallum, R. C., & Briggs, N. E. (2008). Latent
growth curve modeling (Vol. 157). Los Angeles: Sage.

146
RachBeisel, J., Scott, J., & Dixon, L. (1999). Co-occurring severe mental illness and
substance use disorders: A review of recent research. Psychiatric Services, 50,
1427-1434.
Ridgely, M. S., Goldman, H. H., & Willenbring, M. (1990). Barriers to the care of
persons with dual diagnoses: Organizational and financing issues.
Schizophrenia Bulletin, 16, 123-1232.
Riggs, D. S., Rothbaum, B. O., & Foa, E. B. (1995). A prospective examination of
symptoms of Posttraumatic Stress Disorder in victims of nonsexual assault.
Journal of Interpersonal Violence, 10, 201-214.
Rothbaum, B. O., Meadows, E. A., Resick, P. A., & Foy, D. W. (2000). Cognitive-
behavioral therapy. In E. B. Foa, T. M. Keane & M. J. Friedman (Eds.),
Effective Treatments for PTSD (pp. 60-83). Oxford: Guilford.
Salasin, S. E. (2005). Evolution of women's trauma-integrated services at the Substance
Abuse and Mental Health Services Administration. Journal of Community
Psychology, 33, 379-393.
Saunders, B. E., Villeponteaux, J. A., Lipovsky, J. A., Kilpatrick, D. G., & Veronen, L.
(1992). Child sexual assault as a risk factor for mental disorders among women.
A community survey. Journal of Interpersonal Violence, 7, 189-204.
Schnurr, P. P., & Green, B. L. (2004). Trauma, PTSD, and health outcomes. Advances,
20, 18-30.
Schwarz, G. (1978). Estimating the dimension of a model. Annals of Statistics, 6, 461-
464.
Shelby, R. A., Golden-Kreutz, D. M., & Andersen, B. L. (2005). Mismatch of
Posttraumatic Stress Disorder (PTSD) symptoms and DSM-IV symptom clusters
in a cancer sample: Exploratory factor analysis of the PTSD Checklist-Civilian
Version. Journal of Traumatic Stress, 18, 347-357.
Simms, L. J., Watson, D., & Doebbeling, B. N. (2002). Confirmatory factor analyses of
posttraumatic stress symptoms in deployed and nondeployed veterans of the
Gulf War. Journal of Abnormal Psychology, 111, 637-647.
Simpson, D. D., Joe, G. W., & Rowan-Szal, G. A. (1997). Drug abuse treatment
retention and process effects on follow-up outcomes. Drug and Alcohol
Dependence, 47(3), 227-235.
SPSS. (2002). SPSS for Windows, Release 11.5. Chicago: SPSS.

147
Stark, M. J. (1992). Dropping out of substance abuse treatment: A clinically oriented
review. Clinical Psychology Review, 12(1), 93-116.
Steiger, J. H. (1990). Structural model evaluation and modification: An interval
estimation approach. Multivariate Behavioral Research, 25, 173-180.
Stein, M. B., & Kennedy, C. M. (2001). Major depressive and post-traumatic stress
disorder comorbidity in female victims of intimate partner violence. Journal of
Affective Disorders, 66, 133-138.
Stewart, S. H., Conrod, P. J., Pihl, R. O., & Dongier, M. (1999). Relations between
posttraumatic stress symptom dimensions and substance dependence in a
community-recruited sample of substance-abusing women. Psychology of
Addictive Behaviors, 13, 78-88.
Stewart, S. H., Pihl, R. O., Conrod, P. J., & Dongier, M. (1998). Functional associations
among trauma, PTSD, and substance-related disorders. Addictive Behaviors, 23,
797-812.
Substance Abuse and Mental Health Services Administration. (2007). May patients
have co-occurring mental and substance abuse disorders - Both must be
addressed for successful treatment. Retrieved May 27, 2007, 2007, from
http://162.99.3.50/news/newsreleases/050131nr_TIP42.htm
Taylor, S. F., Kuch, K., Koch, W. J., Crockett, D. J., & Passey, G. (1998). The structure
of posttraumatic stress symptoms. Journal of Abnormal Psychology, 107, 154-
160.
Terr, L. (1991). Childhood traumas: An outline and overview. American Journal of
Psychiatry, 148, 10-20.
Timko, C., & Moos, R. H. (2002). Symptom severity, amount of treatment, and 1-year
outcomes among dual diagnosis patients. Administration and Policy in Mental
Health, 30, 35-54.
Tjaden, P., & Thoennes, N. (2000). Prevalence and consequences of male-to-female and
female-to-male intimate partner violence as measured by the National Violence
against Women Survey. Violence against Women, 142, 142-161.
U.S. Department of Health and Human Services. (1999). Mental health: A report of the
surgeon general. Rockville, MD: U.S. Department of Health and Human
Services.

148
van der Hart, O., Nijenhuis, E. R. S., & Steele, K. (2006). The Haunted Self: Structural
Dissociation and the Treatment of Chronic Traumatization. New York: Norton.
van der Kolk, B. A. (1996a). The body keeps score: Approaches to the psychobiology
of post-traumatic stress disorder. In B. A. van der Kolk, A. C. McFarlane & L.
Weisaeth (Eds.), Traumatic Stress: The effects of overwhelming experience on
mind, body and society (pp. 214-241). New York: Guilford Press.
Van der Kolk, B. A. (1996b). The complexity of adaptation to trauma: Self-regulation,
stimulus discrimination, and characterological development. In B. A. Van der
Kolk, A. C. McFarlane & L. Weisaeth (Eds.), Traumatic Stress: The Effects of
Overwhelming Experience on Mind, Body, and Society (pp. 182-213). New
York: Guildford Press.
van der Kolk, B. A., McFarlane, A. C., & Weisæth, L. (1996). Traumatic Stress: The
Effects of Overwhelming Experience on Mind, Body, and Society. New York:
Guilford.
Watkins, K. E., Burnam, A., Kung, F.-Y., & Paddock, S. (2001). A national survey of
care for persons with co-occurring mental and substance use disorders.
Psychiatric Services, 52, 1062-1068.
Wilsnack, S. C., Vogeltranx, N. D., Klassen, A. D., & Harris, R. (1997). Childhood
sexual abuse and women's substance abuse: National survey findings. Journal
of Studies on Alcohol, 58, 264-271.
Witteveen, A. B., Van der Ploeg, E., Bramsen, I., Huizink, A. C., Slottje, P., Smid, T.,
et al. (2006). Dimensionality of the posttraumatic stress response among police
officers and fire fighters: An evaluation of two self-report scales. Psychiatry
Research, 141, 213-228.
Wolfe, J., & Kimerling, R. (1997). Gender issues in the assessment of posttraumatic
stress disorder. In J. P. Wilson & T. M. Keane (Eds.), Assessing Psychological
Trauma and PTSD (pp. 192-238). New York: Guilford.
Yehuda, R., Kahana, B., Schmeidler, J., Southwick, S. M., Wilson, S., & Giller, E. L.
(1995). Impact of cumulative lifetime trauma and recent stress on current
posttraumatic stress disorder symptoms in holocaust survivors. American
Journal of Psychiatry, 152, 1815-1818.
Young, M. S., Hills, H. A., Rugs, D., Peters, R., Moore, K., Woods-Brown, L., et al.
(2004). Integrating Seeking Safety into substance abuse treatment programs.
Paper presented at the 112th annual meeting of the American Psychological
Association, Honolulu, HI.

149
Young, N. K., & Grella, C. E. (1998). Mental health and substance abuse treatment
services for dually diagnosed clients: Results of a statewide survey of county
administrators. Journal of Behavioral Health Services and Research, 25, 83-92.
Zhang, Z., Friedmann, P. D., & Gerstein, D. R. (2003). Does retention matter?
Treatment duration and involvement in drug use. Addiction, 98, 673-684.
Zlotnick, C., Najavits, L. M., Rohsenow, D. J., & Johnson, D. M. (2003). A cognitive-
behavioral treatment for incarcerated women with substance abuse disorder and
posttraumatic stress disorder: Findings from a pilot study. Journal of Substance
Abuse and Treatment, 25, 99-105.
Zweben, J. E. (1996). Psychiatric problems among alcohol and other drug dependent
women. Journal of Psychoactive Drugs, 28, 345-366.

150
APPENDIX A:
SEEKING SAFETY SESSIONS IMPLEMENTED AT LOS ANGLES SITE OF
WCDVS

Table A-1: Description of Seeking Safety Sessions Implemented at Los Angeles Site of
WCDVS

Session # Session Title Skill
Type

Session Summary
8

1 Introduction C, B, I 1) Introduction to the treatment and getting to
know the patient and 2) case management

2 Safety C, B, I Safety is described as the first stage of healing
from both PTSD and substance abuse (SA), and
the foremost guiding principle throughout
treatment. A list of over 80 Safe Coping skills
is provided, and patients explore what safety
means to them.

3 Detaching
from
Emotional
Pain/
Grounding

B A powerful strategy known as “grounding” is
reviewed to help patients detach from emotional
pain. The goal is to shift attention toward the
external world, away from negative feelings.

4 & 5 PTSD, Taking
Back Your
Power, Parts 1
& 2
C Session covers 1) what is PTSD; 2) the link
between PTSD and SA; 3) using compassion to
take back your power; and 4) long-term PTSD
problems. In each of these the goal is to
provide information as well as a compassionate
understanding of the disorder (the key to
“taking back your power”).

Notes. Most sessions focused on one type of coping skills: C = cognitive skills; B =
behavioral skills; I = interpersonal skills.

8
Descriptions of each session are reproduced wholly or adapted from session introductions in the
Seeking Safety manual (Najavits, 2004).

151
Table A-1: Continued

Session # Session Title Skill
Type

Summary
6 & 7 When
Substances
Control You,
Parts 1 & 2
C Handouts are provided to cover eight topics: 1)
what is SA; 2) how SA prevents healing from
PTSD; 3) choosing a way to give up substances;
4) an imaginary exercise to prepare realistically
for giving up substances; 5) coping with mixed
feelings about giving up substances; 6) self-
understanding of substance use; 7) self-help
groups; and 8) common questions about PTSD
and SU.

10 Asking for
Help
I Each of the disorders – PTSD and SA – leads to
problems in asking for help. This topic
encourages women to become aware of their
need fro help, and provides guidance in how to
do so effectively.

11 Taking Good
Care of You
B Women are guided to explore how well they
take care of themselves, using a questionnaire
listing specific behaviors. They are asked to
take immediate action to improve at least one
self-care problem.

12 & 13 Healing from
Anger, Parts 1
& 2
I Anger is explored as a valid feeling that is
inevitable in recovery from PTSD and SA.
Guidelines for working with both constructive
and destructive forms of anger are offered.

14 Compassion C This topic guides women to replace destructive
self-talk with compassionate self-talk. They are
taught that only a loving stance toward the self
produces lasting changes.

Notes. Most sessions focused on one type of coping skills: C = cognitive skills; B =
behavioral skills; I = interpersonal skills.

152
Table A-1: Continued

Session # Session Title Skill
Type

Summary
15 Red & Green
Flags
B Women are guided to 1) identify signs of
danger and safety (“red and green flags”) for
PTSD and SA, and 2) create a safety plan.

16 Honesty I Women are encouraged to explore the role of
honesty in recovery and to role-play specific
situations. Issues relevant to the topic include
the cost of dishonesty, when is it safe to be
honest, and what if the other person does not
accept honesty.
17 Integrating the
Split Self
C Women are guided to recognize the internal
splits inherent in both PTSD and SA, and to
explore ways to integrate them for recovery.

18 Commitment B Women are encouraged to explore the role of
commitments in their lives, to learn creative
strategies for keeping commitments, and to
identify feelings that get in the way.

19 & 20 Creating
Meaning,
Parts 1 & 2
C The topic explores the meaning clients create –
with particular attention to assumptions specific
to PTSD and SU. Women are encouraged to
compare meanings that are harmful versus
healing in recovery.

21 Self-
Nurturing
B This topic seeks to inspire women to increase
pleasurable activities. Safe self-nurturing is
distinguished from unsafe self-nurturing.

22 Health
Relationships
I This topic explores healthy and unhealthy
beliefs about relationships.

Notes. Most sessions focused on one type of coping skills: C = cognitive skills; B =
behavioral skills; I = interpersonal skills.

153
Table A-1: Continued

Session # Session Title Skill
Type

Summary
23 & 24 Setting
Boundaries,
Part 1 & 2
I Boundary problems are described in two forms:
too much closeness (difficulty saying “no” in
relationships) and too much distance (difficulty
saying “yes” in relationships). Ways to set
healthy boundaries are described.

25 Discovery C Women are encouraged to use cognitive
techniques to find out if their beliefs are true,
rather than “staying stuck.” They are provided
with ways to discover (e.g., ask others, “try it
and see,” “predict,” and “act as if”) and
prepared for how to cope with negative
feedback.

28 Respecting
Your Time
B Women are asked to explore how they spend
their time, as a way to better understand their
approach to recovery. Women are guided to try
to use their time to its fullest advantage, while
respecting that recovery from PTSD and SA
means that their use of time may be different
from people without these disorders.

29 Community
Resources
I A list of national resources is offered to aid
recovery (including advocacy organizations,
self-help groups, newsletters, and other
nonprofit organizations). Also, guidelines are
offered to help women take a consumer
approach in selecting and evaluating treatments.

30 Life Choices
Game
C, B, I The Life Choices Game is provided as an
entertaining way to review the treatment (i.e.,
women generate safe and realistic coping
strategies to challenging situations).

Notes. Most sessions focused on one type of coping skills: C = cognitive skills; B =
behavioral skills; I = interpersonal skills.

154
Table A-1: Continued

Session # Session Title Skill
Type

Summary
31 Termination C, B, I The final topic encourages women to express
their feelings about the ending of treatment,
discuss what they liked and disliked about it,
and finalize their aftercare plans.

Notes. Most sessions focused on one type of coping skills: C = cognitive skills; B =
behavioral skills; I = interpersonal skills.

155
APPENDIX B:
PSS-SR ITEMS BY SUBSCALE

Table B-1: PSS-SR Items by Subscale

Re-experiencing Items

Avoidance Items
1. Having upsetting thoughts or images
about the traumatic events that come into
your head when you don’t want them to

6. Trying not to think about, talk about
or have feelings about the traumatic
events.
2. Having bad dreams or nightmares
about the traumatic events*

7. Trying to avoid activities, people or
places that remind you of the traumatic
events*
3. Reliving the traumatic events, acting
or feeling as if they were happening
again*

4. Feeling emotionally upset when you
were reminded of the traumatic events

5. Experiencing physical reactions when
you were reminded of the traumatic
events *

Note. * indicates items dropped from final version of scale.

156
Table B-1. Continued.

Numbing Items

Arousal Items
8. Not being able to remember an
important part of the traumatic events

13. Having trouble concentrating

9. Having much less interest or
participating much less often in
important activities

14. Feeling irritable or having fits of
anger.
10. Feeling distant or cut off from
people around you

15. Having trouble falling or staying
asleep
11. Feeling emotionally numb

16. Being overly alert
12. Feeling as if future plans or hopes
will not come true

17. Being jumpy or easily startled*

Note. * indicates items dropped from final version of scale.
157

APPENDIX C.
ITEMS ON UNSAFE EVENTS SCALE

Table C-1: Items on Unsafe Events Scale

Question Stems:
Baseline: Have you ever (been) . . .
6-months and 12-months: In the last six months have you . . .

1. Sent to jail

7. Been homeless

2.. Had serious money problems

8. Had an abortion
3. Separated from your child(ren)
against your will

9. Seen a robbery, a mugging, or an
attack taking place

4. Physically neglected (“For
example, not fed, not properly
clothed, or left to take care of
yourself when you felt you were too
young or ill.”)

10. Ever touched or made to touch
someone else in a sexual way, because
you felt forced in some way or
threatened by arm to yourself or
someone else

5. Ever had sex when you did not
want to in exchange for money,
drugs, or other material goods such as
shelter or clothing

11. Robbed, mugged, or physically, not
sexually, attacked by a stranger or by
someone you did not know well

6. Physically abused or severely
punished by someone you knew well
such as a parent, sibling, boyfriend or
girlfriend

12. Ever have sex because you felt
forced in some way or threatened by
harm to yourself or someone else

Notes. Not all items on this scale are under the direct control of women in this study and
inclusion of such items are not intended to “blame the victim” for stressful and
traumatic experiences. However, as treatment focused on increasing women’s skills to
keep themselves safe, interacting with safer people, in safer environments, and with
enhanced interpersonal skills, it was hypothesized that the frequencies of these types of
events would nevertheless go down.
158

Table C-1: Continued

Question Stems:
Baseline: Have you ever (been) . . .
6-months and 12-months: In the last six months have you . . .

13. Serious accident or accident 15. Been stalked or has anyone ever
threatened to kill or seriously harm you

14. Emotionally abused or neglected

Notes. Not all items on this scale are under the direct control of women in this study and
inclusion of such items are not intended to “blame the victim” for stressful and
traumatic experiences. However, as treatment focused on increasing women’s skills to
keep themselves safe, interacting with safer people, in safer environments, and with
enhanced interpersonal skills, it was hypothesized that the frequencies of these types of
events would nevertheless go down.

Abstract (if available)

Abstract Interpersonal violence directed against girls and women is both widespread and can lead to serious long-term consequences, including the development of co-occurring mental health and substance abuse disorders. This two-part study investigated psychosocial gains following trauma-focused treatment among women with histories of interpersonal violence and co-occurring disorders. The data are drawn from the Los Angeles site of the national Women, Comorbid Disorders and Violence Study

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

Asset Metadata

Creator Mackintosh, Margaret-Anne (author)

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

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Publication Date 08/03/2009

Defense Date 03/23/2009

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

Tag comorbid disorders,co-occurring disorders,interpersonal violence,OAI-PMH Harvest,outcome study,posttraumatic stress,Psychology,PTSD,Seeking Safety,trauma,Treatment,Women

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Gatz, Margaret (committee chair), Hennigan, Karen (committee member), McArdle, Jack (committee member), Trickett, Penelope K. (committee member)

Creator Email mackinto@usc.edu,maggi@lachesis.info

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2454

Unique identifier UC169305

Identifier etd-Mackintosh-2635 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-181067 (legacy record id),usctheses-m2454 (legacy record id)

Legacy Identifier etd-Mackintosh-2635.pdf

Dmrecord 181067

Document Type Dissertation

Rights Mackintosh, Margaret-Anne

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