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