Journal of Family Psychology
2009, Vol. 23, No. 4, 531–539
© 2009 American Psychological Association
0893-3200/09/$12.00
DOI: 10.1037/a0015877
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Family Functioning Predicts Outcomes for Veterans in Treatment for
Chronic Posttraumatic Stress Disorder
Lynette Evans and Sean Cowlishaw
Malcolm Hopwood
La Trobe University
University of Melbourne
A longitudinal framework was used to examine the competing hypotheses of (a) whether
family functioning predicts changes in posttraumatic stress disorder (PTSD) symptoms or (b)
whether PTSD symptoms predict changes in family functioning. Veterans (N ⫽ 311)
admitted to a treatment program completed a series of questionnaires at 3 time points: at
intake, from intake to completion of a treatment program, and at the 6-month follow-up.
Alcohol use and general mental health symptoms were also measured at intake. A crosslagged panel model using structural equation modeling analyses indicated that family functioning was a moderate predictor of PTSD symptoms at posttreatment and at the 6-month
follow-up. PTSD was not a significant predictor of family functioning across time and alcohol
use, and general mental health symptoms did not affect the overall findings. Further analyses
of PTSD symptom clusters indicated that the avoidance symptom cluster was most strongly
related to family functioning. Targeting family relationships for treatment may be important
in the future for veterans with PTSD.
Keywords: posttraumatic stress, veterans, family functioning
Theoretical models examining the connection between
stress, trauma, and family functioning (e.g., Figley, 1983;
Nelson Goff & Smith, 2005) have consistently emphasized
the bidirectional nature of these interactions, whereby family members mutually influence one another. On one hand,
family functioning may alleviate the impacts of trauma on
an individual. On the other hand, stress crossover may
occur, in which case stress is transmitted and other family
members are affected (Westman, Vinokur, Hamilton &
Roziner, 2004). The complex nature of family outcomes
resulting from one family member being traumatized was
highlighted by Figley’s (1983) early work with combat
veterans and their families. Despite the severe impacts of
trauma on the individual, Figley identified four means by
which the family system potentially altered the traumatic
reaction and reduced posttraumatic stress disorder (PTSD)
symptoms. He argued that close family members (a) identified symptoms, (b) confronted the problem, (c) assisted
with the process of verbally expressing elements of the
trauma, and (d) assisted with resolving the trauma-induced
conflicts. A more recent theoretical approach by Nelson
Goff and Smith (2005) addressed similar questions of stress
and trauma in the family. They emphasized the bidirectional
nature of the interactions that occur among the individuals’
reaction to trauma, family members’ reactions, and processes within the couple and family system. As a result, at
all levels of the system (i.e., individual, couple, or family),
either maladaptive or adaptive responses may occur.
Regardless of the presence of theoretical models postulating how individual and family reactions to trauma mutually affect one another, the empirical evidence on how
family functioning affects PTSD, or vice versa, has been
limited. Studies have largely failed to move beyond examining associations between the experiences of trauma, the
resultant PTSD developed by an individual, and the functioning of others within the individual’s close network.
Nonetheless, these studies have consistently identified associations, although not causality, between PTSD symptoms and a range of outcomes characteristic of family
dysfunction and distress (Evans, McHugh, Hopwood, &
Watts, 2003; MacDonald, Chamberlain, Long, & Flett,
1999; Riggs, Byrne, Weathers, & Litz, 1998; Savarese,
Suvak, King, & King, 2001; Westerlink & Giarranto, 1999).
Partners living with veterans are also prone to experiencing
mental health problems (Dirkzwager, Bramsen, Ader, &
van der Ploeg, 2005). Collectively, these studies have indicated that PTSD was consistently associated with a range of
problematic outcomes for families or partners of veterans.
A clear gap in this literature is the lack of prospective
studies and the limited number of longitudinal investigations into the causal relationship between PTSD and family
functioning. Instead, researchers have had a tendency to
imply that causality exists, generally in the direction of
PTSD affecting family dysfunction (Cook, Riggs, Thompson, Coyne, & Sheikh, 2004; Dirkzwager et al., 2005;
Lynette Evans and Sean Cowlishaw, School of Psychological
Science, La Trobe University, Bundoora, Victoria, Australia; Malcolm Hopwood, Veterans Psychiatry Unit, Heidelberg Repatriation Hospital, Austin Health, University of Melbourne, Heidelberg, Victoria, Australia.
Correspondence concerning this article should be addressed to Lynette
Evans, School of Psychological Science, La Trobe University, Bundoora
3086, Victoria, Australia. E-mail: l.evans@latrobe.edu.au
531
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
532
EVANS, COWLISHAW, AND HOPWOOD
Nelson Goff, Crow, Reisbig, & Hamilton, 2007). Despite
the importance of examining how PTSD and family functioning mutually influence one another in a naturalistic
environment, the issue we wanted to address was whether
family functioning influenced treatment outcomes for veterans with PTSD. Evidence exists in other research areas
suggesting that family functioning may predict recovery
from psychological disorders (Whisman, Uebelacker, &
Bruce, 2006). A number of longitudinal studies on depressive disorders also support family functioning predicting
course of illness during and after treatment (Miller et al.,
1992; Moos, Cronkite, & Moos, 1998a, 1998b). Although
the research to date is consistent with the notions that PTSD
predicts family functioning and that family functioning predicts mental health outcomes, including PTSD symptomatology, over time, there is little direct evidence available to
support these claims.
Current Study
Treatment of family and couples has been identified as a
potential means of enhancing effectiveness of treatment
programs for veterans (Sherman, Zanotti, & Jones, 2005),
but no studies, to our knowledge, have considered the
contribution of family functioning in treatment outcomes
for veterans with chronic PTSD. Although a treatment program was involved in the current instance, the focus of this
study was not on the treatment but rather on the likelihood
of family functioning playing an additional role in outcomes
from intake to posttreatment and posttreatment to followup. Given the theoretical perspectives on the interactional
nature of family functioning and PTSD symptoms (Nelson
Goff & Smith, 2005), we expected that (a) family functioning would make a unique contribution to predicting improvement in PTSD symptoms for veterans after completion
of a treatment program and (b) that PTSD symptoms would
also make a unique contribution and have a reciprocal
relationship with family functioning over the course of
treatment. We took measures at Time 1 (intake), Time 2
(posttreatment), and Time 3 (6-month follow-up). Given the
risk that veterans in this population would have comorbid
symptoms, such as alcohol abuse and other mental health
problems, we controlled for these variables in a follow-up
analysis.
Method
Participants
The current sample, at Time 1, included 489 Australian
veterans entering a PTSD treatment program at the Veterans
Psychiatry Unit at the Austin and Repatriation Medical
Centre (Melbourne, Victoria, Australia). Of the original 489
cases, 177 contained more than 25% missing data, largely
reflecting attrition from the study at the second or third
wave, and we also removed 1 case in which the participant
was female. This left a final sample of 311 for the primary
analyses.
Veterans included in the final sample reported a mean age
of 19.30 years (SD ⫽ 2.20 years) when they joined the army
and a mean of 6.48 years (SD ⫽ 7.49 years) military service.
The group was fairly evenly split between soldiers who
were drafted through National Service into the military (n ⫽
158) and regular soldiers (n ⫽ 153), with some missing
data. The mean age of veterans at the time of treatment was
52.10 years (SD ⫽ 4.74), with an age range of 26 to 77
years. The majority of veterans were exposed to trauma
during the war in Vietnam, but younger veterans, who were
a minority in this sample (1%), had been on peacekeeping
missions. Veterans’ trauma histories were not collated or
available for analyses, although a one-item question about
previous trauma suggested that approximately one third
reported experiencing trauma other than that related to
military events.
Veterans were all diagnosed with PTSD through the use
of the Clinician-Administered PTSD Scale (CAPS) diagnostic interview schedule (Weathers, Keane, & Davidson,
2001). Veterans were assessed by psychiatric registrars who
were trained in the administration of the CAPS before
intake. There are 17 items related to symptoms of PTSD,
and the clinician scores the patient on a scale ranging from
0 to 4 in terms of frequency and intensity of symptoms.
Scores of veterans in this study indicated they had a mean of
78.98 (SD ⫽ 15.75), with a range of 31 to 119. Scores of 40
to 59 are considered evidence of PTSD, and scores within
the range of 60 to 79 are considered evidence of severe
PTSD (Weathers et al., 2001). Veterans reported a mean
PTSD symptom duration of 26.71 years (SD ⫽ 9.07). Veterans were also diagnosed with a range of other comorbid
psychiatric symptoms, with the majority meeting the criteria
for major depressive disorder (n ⫽ 123) or alcohol dependence and abuse (n ⫽ 86) and a small number meeting the
criteria for various anxiety, depressive, and dissociative
disorders.
At intake into the program, 248 (79.7%) veterans reported being married, 14 were in committed relationships
(4.5%), and 36 were separated or divorced (11.6%). The
remainder of the veterans represented small groups who
were widowed, living separately, or never married. By
posttreatment, 246 were married, 13 were in committed
relationships, 36 were separated or divorced, and the remainder were widowed, living separately, or never married.
By the 6-month follow-up, the figures were similar, with
247 reporting being married, 14 in committed relationships,
33 separated or divorced, and the remainder widowed or
single. These figures indicate that veterans remained stable
in terms of their relationship status across the three measurement periods.
Procedure
Procedures complied with ethical guidelines outlined by
the National Health and Medical Research Council. On
admission to the PTSD Treatment Program at the Veterans
Psychiatry Unit, veterans completed a number of questionnaires related to their family functioning, PTSD symptoms,
and psychological health. The veterans were then engaged
in a 12-week treatment program before completing ques-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
FAMILY FUNCTIONING PREDICTS PTSD SYMPTOMS
tionnaires at posttreatment and again at the 6-month followup. The treatment program was a cognitive– behavioral program with recommended evidence-based components such
as exposure, anger management, anxiety management, alcohol withdrawal, problem solving, and management of
depression (Australian Centre for Posttraumatic Mental
Health, 2007). The program involved a combination of
inpatient and outpatient phases. There was an intensive
(inpatient or residential) program, which consisted of a
4-week intensive phase (5 days per week) and an 8-week
outpatient phase (1 day per week). Partners were also seen
weekly over the 3-month phase of the program in group
sessions. All patients were seen by a psychiatrist and, if
necessary, prescribed a range of medications. The first line
of treatment was with antidepressant medication, usually
the selective serotonin reuptake inhibitors, with sertraline
and paroxetine the most widely used. A significant proportion of patients were also prescribed benzodiazepines, particularly as a hypnotic. A small proportion of patients were
also prescribed other agents to help arousal, such as atypical
antipsychotics.
Measures
Posttraumatic stress disorder. The PTSD Checklist
Military Version (PCL-M; Weathers, Litz, Herman, &
Keane, 1993) was used to assess the veterans’ PTSD symptoms. This scale is a self-report rating scale, which is useful
in both diagnosing combat-related PTSD and measuring
symptom severity. Participants indicate the degree to which
they experienced the 17 Diagnostic and Statistical Manual
of Mental Disorders (4th ed.; American Psychiatric Association, 1994) symptoms in the past month on a scale
ranging from 1 (not at all) to 5 (extremely). Scores range
from 17 to 85, with a cutoff of 50 indicating a PTSD
diagnosis. The items can be divided into the three main
symptom clusters of PTSD: intrusion (Items 1–5), avoidance (Items 6 –12), and arousal (Items 13–17). The PCL-M
has a test–retest reliability of .96. The reliability score in
this study (Cronbach’s ␣ ⫽ .89) was high.
Family functioning. The McMaster Family Assessment
Device (FAD-12; Epstein, Baldwin, & Bishop, 1983) 12item General Functioning Scale in this study was used to
assess family functioning. The self-report questionnaire
contains a series of statements relating to family interactions
(i.e., partner and children). Questions are responded to on a
4-point scale ranging from 1 (strongly agree) to 4 (strongly
disagree). High scores on the scale indicate unhealthy family functioning and low scores indicate healthy family functioning. The General Functioning Scale has a reported internal reliability of .83 to .86 and correlates from .85 to .88
with the other subscales of the FAD-12 (Kabacoff, Miller,
Bishop, Epstein, & Keitner, 1990). In this sample, the mean
item score on the General Functioning Scale was 2.50
(SD ⫽ 0.53) at intake, indicating that the group was reporting high levels of family dysfunction (Fisher & Corcoran,
2007). The reliability score in this study was high (Cronbach’s ␣ ⫽ .85).
533
Mental health symptoms. The General Health Questionnaire (GHQ-28; Goldenberg & Hillier, 1979) was used to
measure the level of depression, anxiety, and social functioning of veterans in this study. Questions are rated on a
4-point scale ranging from 0 (not at all) to 3 (much more
than usual). The reliability score in this study was high
(Cronbach’s ␣ ⫽ .89).
Alcohol use. The Alcohol Use Disorders Identification
Test (AUDIT; Saunders, Aasland, Babor, de la Fuente, &
Grant, 1993) was used to identify veterans who were at risk
of or who were currently experiencing alcohol problems.
Total scores range from 0 to 40; scores of 8 or higher
identify those at risk of or currently experiencing alcohol
problems. The sensitivity of the AUDIT in detecting alcohol
problems is around .90. The reliability score in this study
was high (Cronbach’s ␣ ⫽ .93).
Statistical Analyses
Overview. We used structural equation modeling
(SEM) in AMOS to analyze the data. SEM has advantages
over alternative analytic techniques (e.g., cross-lagged correlation analysis or multiple regression) that make it preferable for analyzing longitudinal relationships. The chisquare statistic was the primary index of fit used in this
study, and a nonsignificant value indicated a good-fitting
model. Other criteria for evaluating model fit based on the
recommendations of Hu and Bentler (1999) were used in the
current study and included (a) standardized root-meansquare residual (SRMR) less than .08; (b) root-mean-square
error of approximation (RMSEA) less than .06; (c) confirmatory fit index (CFI) greater than .95; and (d) TuckerLewis Index (TLI) greater than .95. We used item parcels as
manifest indicators of latent variables (Little, Cunningham,
Shahar, & Widaman, 2002). Item parceling is a suitable
strategy for representing latent variables when constructs
are unidimensional (Little et al., 2002). In the current study,
exploratory factor analysis supported the presence of a
strong primary factor underlying the FAD-12 (whole scale)
and PCL-M (subscale) items at all times. More specifically,
the majority of the variance in each item pool was captured
by a first factor, and the ratio of the first eigenvalue to the
second was approximately equal to, or greater than, 2.5 to 1
(Hall, Snell, & Foust, 1999). On the basis of Little et al.’s
(2002) recommendations, we created three parcels from the
FAD-12 items (e.g., Parcel 1 was created by summing Items
1, 2, 3, and 4). The PCL-M item parcels were created by
summing item scores corresponding to the PCL-M (i.e.,
intrusion, avoidance, and hyperarousal) subscales, consistent with the “isolated item parceling” approach (Bandalos,
2002). These subscales demonstrated satisfactory reliability,
and the loadings of family functioning and PCL-M item
parcels on their respective latent variables were all positive
and significant (p ⬍ .001) with a median loading of .88.
We evaluated structural relationships among latent variables using a cross-lagged panel design (see Martens &
Haase, 2006). Martens and Haase (2006) outlined a process
for testing cross-lagged models in SEM that involves testing
the fit of a baseline model with only autoregressive effects
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
534
EVANS, COWLISHAW, AND HOPWOOD
(i.e., no cross-lagged pathways) and then comparing this
against alternative models. These include (a) a model with
autoregressive effects and family functioning predicting
subsequent PTSD, (b) a model with autoregressive effects
and PTSD predicting subsequent family functioning, and (c)
a fully cross-lagged model with autoregressive effects and
both cross-lagged pathways. Given that alternative models
are nested within the baseline model, the chi-square change
statistic can be used to compare the fit of each model before
the parameters of the best-fitting model are interpreted
(Anderson & Gerbing, 1988). We conducted supplementary
analyses to examine the relations between family functioning and specific PSTD symptom clusters. In particular, we
ran a series of three sequential cross-lagged panel models
with the PTSD symptoms of each subscale (i.e., intrusion,
avoidance and hyperarousal) entered independently. A further supplementary analysis was run with the original crosslagged model using single indicator variables for alcohol
use and general mental health symptoms to control for other
factors comorbid with PTSD.
Results
Descriptive Analyses
Independent group t tests enabled us to screen for the
effects of missing data. Results showed that the sample with
more than 25% missing data did not differ from the remaining cases on age, family functioning at intake, or PTSD at
intake. As such, there was no evidence to suggest that cases
with large amounts of missing data differed on family
dysfunction or PTSD from the remaining cases, and these
cases were excluded from the analysis. Of the remaining
311 cases, 85 still demonstrated some missing data, and we
used the expectation–maximization algorithm (Graham,
Hofer, Donaldson, MacKinnon, & Schafer, 1997) to replace
these missing values.
We calculated descriptive statistics for measures of PTSD
and family functioning across all waves and for the AUDIT
and GHQ-28 at Time 1. Given that we used item parceling
strategies to represent latent variables in the SEM analyses,
we supply these scale scores here for descriptive purposes
only. Means and standard deviations for all study variables
are presented in Table 1. A repeated measures analysis of
variance, with repeated contrasts between Time 1 to Time 2
and Time 2 to Time 3, indicated that there was a significant
overall difference on the PCL-M scores, F(2, 309) ⫽ 71.54,
p ⬍ .001, with significant change recorded between Time 1
and Time 2, F(1, 310) ⫽ 69.77, p ⬍ .001, and Time 2 and
Time 3, F(1, 310) ⫽ 19.45, p ⬍ .001. There was no
significant change on the FAD-12 scores overall, F(2,
309) ⫽ 2.82, p ⫽ .06, across the three time points. Table 2
displays correlations between all study variables.
Cross-Lagged Panel Analysis
As mentioned previously, we tested four alternative models in the main cross-lagged analysis, including (a) a baseline autoregressive model; (b) a model with autoregressive
effects and pathways from PTSD to family functioning; (c)
a model with autoregressive effects and pathways from
family functioning to PTSD; and (d) a fully cross-lagged
model. As recommended in the literature (e.g., Kline, 2005),
we allowed within-time residuals associated with endogenous variables to covary, as well as error terms for corresponding manifest variables measured at different times.
Preliminary data screening indicated that several variables
were non-normally distributed, and Mardia’s normalized
multivariate kurtosis statistic also indicated multivariate
non-normality within the data. Given that the maximum
likelihood estimation method is based on assumptions of
multivariate normality, we calculated the Bollen-Stine bootstrap p to compensate for biased probability estimates. Fit
indices for each model are presented in Table 3.
As can be seen from Table 3, the autoregressive model
showed a reasonable fit to the data, as reflected in RMSEA,
CFI, and TLI values all within the desired range. However,
the chi-square statistic was significant (suggesting a lack of
exact fit to the data), and the SRMR was only marginally
acceptable. The PTSD 3 Family model also showed a
reasonable fit to the data, although this model did not fit the
data significantly better, ⌬2(2) ⫽ 3.73, p ⬎ .05, than the
baseline model. The Family 3 PTSD model was tested next
and also fit the data well, as reflected in SRMR, RMSEA,
CFI, and TLI values all within the desired range. Further-
Table 1
Means and Standard Deviations of All Study Measures
T1
T2
T3
Variable
M
SD
M
SD
M
SD
FAD-12 mean of items
PCL-M (total)
Intrusion
Avoidance
Hyperarousal
AUDIT
GHQ-28
2.50
65.55
17.93
26.76
20.86
15.85
42.08
0.53
10.46
4.19
4.92
3.20
12.52
15.62
2.47
60.60
16.07
24.91
19.61
0.54
11.39
4.09
5.33
3.60
2.43
58.33
15.44
24.05
18.84
0.57
11.22
4.02
5.25
3.70
Note. FAD-12 ⫽ Family Assessment Device, 12-item General Functioning Scale; PCL-M ⫽
PTSD Checklist Military Version; AUDIT ⫽ Alcohol Use Disorders Identification Test; GHQ-28 ⫽
General Health Questionnaire; T1 ⫽ Time 1 (intake); T2 ⫽ Time 2 (posttreatment); T3 ⫽ Time 3
(6-month follow-up).
FAMILY FUNCTIONING PREDICTS PTSD SYMPTOMS
535
Table 2
Intercorrelations Between the Family Functioning, Posttraumatic Stress Disorder Symptoms, Alcohol Use, and General
Mental Health Symptoms Across Three Time Points
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Variable
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
T1 FAD-12
T1 PCL-M
T1 PCL-M Intrusion subscale
T1 PCL-M Avoidance subscale
T1 PCL-M Hyperarousal subscale
T2 FAD-12
T2 PCL-M
T2 PCL-M Intrusion subscale
T2 PCL-M Avoidance subscale
T2 PCL-M Hyperarousal subscale
T3 FAD-12
T3 PCL-M
T3 PCL-M Intrusion subscale
T3 PCL-M Avoidance subscale
T3 PCL-M Hyperarousal subscale
Alcohol Identification Test
General Health Questionnaire
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
—
.34
.18
.40
.26
.60
.31
.19
.35
.25
.60
.37
.22
.40
.32
.10
.30
—
.83
.89
.81
.19
.55
.50
.47
.47
.17
.51
.47
.45
.41
.07
.55
—
.58
.52
.18
.44
.54
.30
.33
.04
.42
.54
.28
.29
.05
.40
—
.62
.23
.47
.36
.49
.35
.23
.44
.33
.47
.30
.06
.52
—
.13
.49
.37
.38
.57
.16
.45
.32
.37
.48
.08
.47
—
.42
.27
.46
.34
.76
.40
.24
.43
.34
.10
.23
—
.84
.90
.86
.40
.67
.56
.58
.61
.09
.40
—
.40
.48
.25
.57
.62
.40
.48
.11
.34
—
.70
.46
.60
.41
.62
.50
.04
.37
—
.31
.60
.44
.48
.66
.09
.33
—
.51
.29
.56
.44
.15
.22
—
.82
.89
.87
.17
.37
—
.56
.62
.14
.33
—
.69
.12
.31
—
.18
.32
—
.19
—
Note. Pearson’s correlations between .12 and .14 were significant at p ⬍ .05; correlations greater than .15 were significant at p ⬍ .01.
FAD-12 ⫽ Family Assessment Device, 12-item General Functioning Scale; PCL-M ⫽ PTSD Checklist Military Version; T1 ⫽ Time 1
(intake); T2 ⫽ Time 2 (posttreatment); T3 ⫽ Time 3 (6-month follow-up).
more, this model provided a significantly better fit to the
data than did the autoregressive model, ⌬2(2) ⫽ 16.94,
p ⬍ .001. Finally, the fully cross-lagged model was found to
provide good fit to the data and fit better than the autoregressive model, ⌬2(4) ⫽ 19.81, p ⬍ .001. Given that the
fully-cross lagged model demonstrated the lowest chisquare value, this was compared with the Family 3 PTSD
model. Results showed that the fully cross-lagged model
was not a significantly better fit to the data than the Family
3 PTSD model, ⌬2(2) ⫽ 2.6, p ⬎ .05. In the fully
cross-lagged model, the pathways from PTSD to family
functioning at subsequent times were not statistically significant. Standardized parameter estimates are shown in
Figure 1 (given that the nonsignificant relationship between
PTSD and family functioning was unexpected, beta values
for this pathway are also shown). To decrease complexity,
the measurement model is not displayed. In reference to
Figure 1, R2 values indicated that the stability and crosslagged pathways accounted for 48% of the variance in Time
2 family functioning and 70% of the variance in Time 3
family functioning. Autoregressive and cross-lagged pathways also accounted for 35% of the variance in Time 2
PTSD and 53% of the variance in Time 3 PTSD. Beta
coefficients for the autoregressive pathways were large and
significant. However, of primary interest for this model, the
beta coefficient from Time 1 family functioning to Time 2
PTSD was 0.16, indicating that a 1 standard deviation
increase in Time 1 family functioning predicted a 0.16
standard deviation increase in Time 2 levels of PTSD when
Time 1 values of PTSD were held constant. We found a
similar beta coefficient ( ⫽ 0.17) for the pathway between
Time 2 family functioning and Time 3 PTSD. In contrast,
neither of the pathways from PTSD (at either Time 1 or
Time 2) to subsequent values of family functioning were
statistically significant. Results supported a stronger predictive relationship from family functioning to PTSD than
from PTSD to family functioning. All veterans who were
not in a current partner relationship were removed from the
data, and the model was reanalyzed with this data set (n ⫽
285). There were no changes noted to the overall model,
although there was a slight decrease in the pathway from
FAD-12 scores at intake to PCL-M scores posttreatment
( ⫽ 0.12) and a slight increase ( ⫽ 0.20) in the pathway
Table 3
Model Fit Statistics
Model
Autoregressive
PTSD 3 Family
Family 3 PTSD
Fully cross-lagged
2
ⴱⴱⴱ
212.127
208.398ⴱⴱ
195.180ⴱⴱ
192.314ⴱⴱ
df
SRMR
RMSEA
CFI
TLI
110
108
108
106
.08
.07
.05
.05
.06
.06
.05
.05
.98
.98
.98
.98
.97
.97
.97
.97
Note. SRMR ⫽ standardized root-mean-square residual; RMSEA ⫽ root-mean-square error of
approximation; CFI ⫽ confirmatory fit index; TLI ⫽ Tucker-Lewis index; PTSD ⫽ posttraumatic
stress disorder.
ⴱⴱ
p ⬍ .05. ⴱⴱⴱ p ⬍ .01.
536
EVANS, COWLISHAW, AND HOPWOOD
z3
Family
Functioning
T1
.43**
.72*
-.07
PTSD
T1
.16*
.50**
z1
.80**
Family
Functioning
T2
.45**
.46**
.17*
.07
PTSD
T2
Family
Functioning
T3
.63**
PTSD
T3
z4
z2
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Figure 1. Standardized parameter estimates for the fully cross-lagged model. T1 ⫽ Time 1
(intake); T2 ⫽ Time 2 (posttreatment); T3 ⫽ Time 3 (6-month follow-up); PTSD ⫽ posttraumatic
stress disorder. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01.
from FAD-12 at posttreatment to the PCL-M at the 6-month
follow-up.
Analyses of Subscales and Comorbid Symptoms
We then estimated three cross-lagged models, each containing a separate PCL-M subscale. In each model, the
PCL-M subscale score was specified as a single indicator of
a latent variable with a regression weight of 1 and an error
variance of zero. Table 4 provides the model fit statistics
and parameter estimates for the four cross-lagged pathways
in each model. The fit statistics indicated good fit to the data
in each case. The cross-lagged parameter estimates showed
that the relationships between family functioning and PTSD
symptoms were largest for avoidance symptoms from Time
1 to Time 2 ( ⫽ 0.20) and from Time 2 to Time 3 ( ⫽
0.21).
Comorbid symptoms of alcohol use and general mental
health symptoms, although addressed during treatment, may
have confounded research findings in our study. For instance, one alternative explanation for the observed results
in this study was that effects from family functioning to
PTSD were attributable to shared associations with comorbid problems. To evaluate this possibility, we conducted
supplementary analyses, using additional data with baseline
levels of alcohol problems and general mental health symptoms. Accordingly, we respecified a SEM model with the
inclusion of the AUDIT and the GHQ-28 as single indicator
variables. Table 4 provides the model fit statistics and
parameter estimates for the cross-lagged pathways. The fit
statistics indicated good fit to the data, and the cross-lagged
parameter estimates showed that the relationships from the
FAD-12 to the PCL-M remained significant even after including the AUDIT and the GHQ-28. The beta coefficients
from the AUDIT to the PCL-M were 0.02 (p ⬎ .05) for
Time 2 and 0.10 (p ⬍ .05) for Time 3; from the AUDIT to
the FAD-12, they were 0.01 (p ⬎ .05) for Time 2 and 0.08
(p ⬍ .05) for Time 3. The beta coefficients from the
GHQ-28 to the PCL-M were 0.10 (p ⬎ .05) for Time 2 and
Time 3; from the GHQ-28 to the FAD-12, they were 0.08
(p ⬎ .05) for Time 2 and 0.01 (p ⬎ .05) for Time 3.
Discussion
Our study has demonstrated that after controlling for
prior levels of family dysfunction and PTSD symptoms, higher
levels of family dysfunction were associated with increased
levels of PTSD symptoms over the course of treatment. This
pathway extending from family functioning to PTSD across
the first to the second wave, and thus over the course of
treatment, was repeated from the second to the third wave,
and thus from posttreatment to 6-month follow-up. In contrast, PTSD symptoms did not predict family dysfunction
longitudinally. Therefore, although PTSD symptoms appeared to be affected by treatment (as reflected in declines
in PTSD symptom levels over time on average), when
holding prior levels of PTSD symptoms constant, family
functioning did affect the veterans’ symptoms and thus their
Table 4
Model Fit Statistics and Cross-Lagged Parameter Estimates for the PCL-M Subscales and the Model which Included
AUDIT and GHQ-28 Analyses
Cross-lagged parameter estimates
Model
2
Intrusion
109.17ⴱⴱⴱ
Avoidance
84.80ⴱⴱ
Hyperarousal
81.20ⴱⴱ
AUDIT & GHQ-28 219.75ⴱⴱ
df
37
37
37
130
SRMR RMSEA CFI TLI
.04
.04
.04
.05
.08
.07
.06
.05
.97
.98
.98
.98
.95
.97
.97
.97
T1 FAD-12 3 T1 PCL-M 3 T2 FAD-12 3 T2 PCL-M 3
T2 PCL-M
T2 FAD-12
T3 PCL-M
T3 FAD-12
.11ⴱ
.20ⴱⴱⴱ
.12ⴱ
.16ⴱ
⫺.05
⫺.06
⫺.05
⫺.12
.07
.21ⴱⴱⴱ
.14ⴱ
.16ⴱⴱ
.03
.09ⴱ
.02
.06
Note. SRMR ⫽ standardized root-mean-square residual; RMSEA ⫽ root-mean-square error of approximation; CFI ⫽ confirmatory fit
index; TLI ⫽ Tucker-Lewis index; FAD-12 ⫽ Family Assessment Device, General Functioning Scale; PCL-M ⫽ PTSD Checklist
Military Version; AUDIT ⫽ Alcohol Use Disorders Identification Test; GHQ-28 ⫽ General Health Questionnaire; T1 ⫽ Time 1 (intake);
T2 ⫽ Time 2 (posttreatment); T3 ⫽ Time 3 (6-month follow-up).
ⴱ
p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
FAMILY FUNCTIONING PREDICTS PTSD SYMPTOMS
ability to benefit from treatment. However, PTSD symptoms demonstrated by the veterans did not have significant
effects on their family system, despite the fact that partners
attended a weekly group session.
A key finding from this study was that family functioning
played a role in successful or unsuccessful change for
veterans. This is consistent with theoretical models arguing
for the interactional nature of individual PTSD symptoms
and family functioning (e.g., Nelson Goff et al., 2006;
Nelson Goff & Smith, 2005). The idea of the relational
context being important in individual psychopathology is
not a new concept (e.g., Coyne, 1976). Surprisingly, in our
study we found no relationship from PTSD to family functioning. We expected that there would be a reciprocal relationship, particularly over a treatment program. In particular, we expected that as veterans reported change in their
individual symptoms, this would have reciprocal impacts on
their family system. The current results did not support this
hypothesis. We should note, however, that although PTSD
symptoms reduced, on average, from intake to posttreatment and then from posttreatment to follow-up, the mean
rating on the PCL-M scale remained high, and many veterans still scored within the clinical range. It remains to be
investigated in further research whether the family functioning in veterans with reduced PTSD symptoms after treatment eventually shifts to accommodate the changes made
by the individual. Our time frame of a 6-month follow-up
may have been too short to observe such changes, given that
in this study, the family system had been coping with the
individual PTSD symptoms for long periods of time (an
average of 26 years) and in some cases individuals had
complicated trauma reactions as a consequence of experience with multiple traumas.
There are two possibilities for why family functioning
predicted PTSD symptoms at the end of treatment in this
study. The first possibility is that families who are more
adaptive encourage, through the dynamics in the family
system, positive change in the veteran. Consistent with this
possibility, studies with women with cancer have suggested
that their partner’s ability to communicate effectively had a
major role in promoting positive treatment outcomes for the
women (Manne et al., 2004). The second explanation is that
families with high levels of dysfunction undermine treatment gains. Treatment outcomes for individuals with mental
illness have shown that negative processes within the family
raise the risk of higher relapse (Lopez, Nelson, Snyder, &
Mintz, 1999). These dynamics of families and their impact
on treatment outcomes of the individual need further exploration.
The follow-up analyses in our study indicated that family
functioning predominantly influenced the avoidance symptoms. Veterans with poor family functioning may therefore
be more likely to rely on using withdrawal strategies. This
approach to distress in the family appears to impede their
progress in treatment. Given that it has been illustrated in
the literature that social support plays an important role in
posttrauma experience (Brewin, Andrews, & Valentine,
2000), intervention with veterans with PTSD where the
family or couple display poor interpersonal functioning
537
(Monson, Schnurr, Stevens, & Guthrie, 2004) may be essential for the veteran to progress individually. Family interventions should be targeted as an important area to develop in the future with particular emphasis on the
avoidance strategies. The importance of comorbid symptoms such as alcohol abuse and general mental health symptoms also requires further follow-up. Our analyses were
preliminary in relation to how these symptoms influenced
family functioning and PTSD; however, family functioning,
over and above these comorbid symptoms, significantly
affected change in PTSD symptoms over time. This again
emphasizes the importance of considering family functioning as a treatment option in veterans with PTSD.
Limitations
A feature of this study that has implications for the
findings was the use of a convenience sample. We relied on
investigating veterans who had experienced PTSD symptoms (sometimes for many years) who presented to a hospital setting. This clearly biases the study toward a sample
of veterans with more severe symptoms in terms of PTSD
and other comorbid symptoms. The sample represented a
chronic and problematic group with a specific clinical presentation. We used psychometrically valid instruments to
measure constructs, and we compared participants who
completed questionnaires with those who did not. However,
our measure of family functioning was brief and provided
limited information about specific family dynamics (e.g.,
intimacy and problem solving). Furthermore, we relied on a
moderate sample size and only the veteran’s report of family functioning and PTSD symptoms. This limits the impact
of the findings given that we were interested in the family
system, which is clearly represented by a wider group. The
Monson et al. (2004) study is an indicator of how others,
including partners, may perceive the family and couple
interactions from a different perspective. Future research
should focus on using a larger sample size, incorporating
partners or other family members in the analyses, and extending measures to capture the dynamics of the family
interaction more fully.
Conclusions
The interaction between the intrapersonal and the interpersonal in relation to treatment outcomes indicates potential for fruitful development of treatment programs in the
context of many clinical disorders. There have been recommendations to develop treatment guidelines and incorporate
family systems thinking into mainstream psychiatric practice (Schweitzer, Zwack, Weber, Nicolai, & Hirschenberger, 2007). Work such as our study points the way and
identifies the need to shift treatment focus from an individual orientation and consider family perspectives, at least in
the case of veterans with PTSD, if not in terms of other
clinical populations.
538
EVANS, COWLISHAW, AND HOPWOOD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
References
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC:
Author.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation
modeling practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411– 423.
Australian Centre for Posttraumatic Mental Health. (2007). Australian guidelines for the treatment of adults with acute stress
disorder and posttraumatic stress disorder. Melbourne, Victoria,
Australia: Author.
Bandalos, D. L. (2002). The effects of item parceling on goodnessof-fit and parameter estimate bias in structural equation modeling. Structural Equation Modeling, 10, 78 –102.
Brewin, C. R., Andrews, B., & Valentine, J. D. (2000). Metaanalysis of risk factors for posttraumatic stress disorder in
trauma-exposed adults. Journal of Consulting and Clinical Psychology, 68, 748 –766.
Cook, J. M., Riggs, D. S., Thompson, R., Coyne, J. C., & Sheikh,
J. I. (2004). Posttraumatic stress disorder and current relationship
functioning among World War II ex-prisoners of war. Journal of
Family Psychology, 18, 36 – 45.
Coyne, J. C. (1976). Depression and the response of others. Journal of Abnormal Psychology, 85, 186 –193.
Dirkzwager, A. J. E., Bramsen, I., Ader, H., & van der Ploeg,
H. M. (2005). Secondary traumatization in partners and parents
of Dutch peacekeeping. Journal of Family Psychology, 19, 217–
226.
Epstein, N. B., Baldwin, L. M., & Bishop, D. S. (1983). The
McMaster Family Assessment Device. Journal of Marital and
Family Therapy, 9, 171–180.
Evans, L., McHugh, A. F., Hopwood, M., & Watts, C. (2003).
Chronic posttraumatic stress disorder and family functioning of
Vietnam veterans and their partners. Australian and New Zealand Journal of Psychiatry, 37, 1– 8.
Figley, C. R. (1983). Stress and the family. In H. I. McCubbin &
C. R. Figley (Eds.), Stress and the family (pp. 45). New York:
Brunner/Mazel.
Fisher, J., & Corcoran, K. (Eds.). (2007). Measures for clinical
practice and research: A sourcebook. Oxford, England: Oxford
University Press.
Goldenberg, D. P., & Hillier, V. F. (1979). A scaled version of the
General Health Questionnaire. Psychological Medicine, 9, 135–
145.
Graham, J. W., Hofer, S. M., Donaldson, S. I., MacKinnon, D. P.,
& Schafer, J. L. (1997). Analysis with missing data in prevention
research. In B. J. Kendall, M. Windle, & S. G. West (Eds.), The
science of prevention: Methodological advances from alcohol
and substance abuse research (pp. 325–366). Washington, DC:
American Psychological Association.
Hall, R. J., Snell, A. F., & Foust, M. S. (1999). Item parceling
strategies in SEM: Investigating the subtle effect of unmodeled
secondary constructs. Organizational Research Methods, 2, 233–
256.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes
in covariance structure analysis: Conventional criteria versus
new alternatives. Structural Equation Modeling, 6, 1–55.
Kabacoff, R. I., Miller, I. W., Bishop, D. S., Epstein, N. B., &
Keitner, G. I. (1990). A psychometric study of the McMaster
Family Assessment Device in psychiatric, medical and nonclinical samples. Journal of Family Psychology, 3, 431– 439.
Kline, R. B. (2005). Principles and practice of structural equation
modeling (2nd ed., pp. 237–287) New York: Guilford Press.
Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F.
(2002). To parcel or not to parcel: Exploring the question,
weighing the merits. Structural Equation Modeling, 9, 151–173.
Lopez, S. R., Nelson, K. A., Snyder, K. S., & Mintz, J. (1999).
Attributions and affective reactions of family members and
course of schizophrenia. Journal of Abnormal Psychology, 108,
307–314.
MacDonald, C., Chamberlain, K., Long, N., & Flett, R. (1999).
Posttraumatic stress disorder and interpersonal functioning in
Vietnam veterans: A meditational model. Journal of Traumatic
Stress, 12, 701–707.
Manne, S. L., Ostroff, J., Sherman, M. D., Heyman, R. E., Ross,
S., & Fox, K. (2004). Couples’ support-related communication,
psychological distress, and relationship satisfaction among
women with early stage breast cancer. Journal of Consulting and
Clinical Psychology, 72, 660 – 670.
Martens, M. P., & Haase, R. F. (2006). Advanced applications of
structural equation modeling in counseling psychology research.
Counseling Psychologist, 34, 878 –911.
Miller, I. W., Gabor, K. I., Whisman, M. A., Ryan, C. E., Epstein,
N. B., & Bishop, D. S. (1992). Depressed patients with dysfunctional families: Description and course of illness. Journal of
Abnormal Psychology, 101, 637– 646.
Monson, C. M., Schnurr, P. P., Stevens, S. P., & Guthrie, K. A.
(2004). Cognitive-behavioral couple’s treatment for posttraumatic stress disorder: Initial findings. Journal of Traumatic
Stress, 17, 341–344.
Moos, R. H., Cronkite, R. C., & Moos, B. S. (1998a). Family and
extrafamily resources and the 10-year course of treated depression. Journal of Abnormal Psychology, 107, 450 – 460.
Moos, R. H., Cronkite, R. C., & Moos, B. S. (1998b). The
long-term interplay between family and extrafamily resources
and depression. Journal of Family Psychology, 12, 326 –343.
Nelson Goff, B. S., Crow, J. R., Reisbig, A., & Hamilton, S.
(2007). The impact of individual trauma symptoms of deployed
soldiers on relationship satisfaction. Journal of Family Psychology, 21, 344 –353.
Nelson Goff, B. S., Reisbig, A. M., Bole, A., Scheer, T., Hayes, E.,
Archuleta, K. L., et al. (2006). The effects of trauma on intimate
relationships: A qualitative study with clinical couples. American
Journal of Orthopsychiatry, 76, 451– 460.
Nelson Goff, B. S., & Smith, D. B. (2005). Systematic traumatic
stress: The couple adaptation to traumatic stress model. Journal
of Marital and Family Therapy, 31, 145–157.
Riggs, D. S., Byrne, C. A., Weathers, F. W., & Litz, B. T. (1998).
The quality of intimate relationships of male Vietnam veterans:
Problems associated with posttraumatic stress disorder. Journal
of Traumatic Stress, 11, 87–101.
Saunders, J. B., Aasland, O. G., Babor, T. F., de la Fuente, T. F.,
& Grant, M. (1993). Development of the Alcohol Use Disorders
Identification Test (AUDIT): WHO collaborative project on
early detection of persons with harmful alcohol consumption.
Addiction, 88, 791– 804.
Savarese, V. W., Suvak, M. K., King, L. A., & King, D. W. (2001).
Relationships among alcohol use, hyperarousal, and marital
abuse and violence in Vietnam veterans. Journal of Traumatic
Stress, 14, 717–732.
Schweitzer, J., Zwack, J., Weber, G., Nicolai, E., & Hirschenberger, N. (2007). Family systems psychiatry: Principles, good
practice guidelines, clinical examples, and challenges. American
Journal of Orthopsychiatry, 77, 377–385.
Sherman, M. D., Zanotti, D. K., & Jones, D. E. (2005). Key
elements in couple’s therapy with veterans with combat-related
posttraumatic stress disorder. Professional Psychology, Research
and Practice, 36, 626 – 633.
FAMILY FUNCTIONING PREDICTS PTSD SYMPTOMS
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Weathers, F. W., Keane, T. M., & Davidson, J. R. T. (2001).
Clinician administered PTSD scale: A review of the first ten
years of research. Depression and Anxiety, 13, 132–156.
Weathers, F. W., Litz, B. T., Herman, D. S., & Keane, T. M.
(1993, October). The PTSD Checklist (PCL-M): Reliability,
validity and diagnostic utility. Paper presented at the International Society of Traumatic Stress Studies, San Antonio,
Texas.
Westerlink, J., & Giarranto, L. (1999). The impact of posttraumatic stress disorder on partners and children of Australian
Vietnam veterans. Australian and New Zealand Journal of Psychiatry, 33, 841– 847.
539
Westman, M., Vinkour, A. D., Hamilton, V. L., & Roziner, I.
(2004). Crossover of marital dissatisfaction during military
downsizing among Russian army officers and their spouses.
Journal of Applied Psychology, 89, 769 –779.
Whisman, M. A., Uebelacker, L. A., & Bruce, M. L. (2006).
Longitudinal association between marital dissatisfaction and alcohol use disorders in a community sample. Journal of Family
Psychology, 20, 164 –167.
Received July 10, 2008
Revision received February 27, 2009
Accepted February 28, 2009
䡲